The "You All Should Know This Already" Warning: I am posting this in SC2 General because it is not a strategy thread. I do not play SC2 regularly and I am not interested in strategy right now. I am not complaining about balance - I know jack about SC2 balance and while I have strong opinions about BW balance, I don't have the standing to express them too vocally, this isn't a balance thread, and (most importantly) this isn't the BW forum - but balance will enter into the discussion a little bit. This does not give you permission to QQ. If you have something to say, please make sure you're actually addressing my analysis.
Introduction
I've seen the DPS question pop up a couple times recently in posts in threads that interested me: GhostFall's twothreads on game design and what he essentially alleges is Blizzard trying too hard, one might say "overbalancing" the game. (That was a summary - go read the threads if you want to debate his points. Don't argue it here.) As a result I thought I would run some basic averages to see how the two games compared. Yes - I am essentially just averaging DPS and comparing the two games.
Methodology
While I started by comparing simple DPS average, this seemed oversimplified. The measure I eventually went with was DPS compared to the average hit points of all units in the study. This has its limitations and inaccuracies, of course, but it seemed to give the most reasonable numbers for comparison within each game - further thoughts will be detailed later.
Even this would have been a fairly complicated way to analyze the entire game, so I decided to make several simplifying assumptions which may potentially throw my numbers way off, so it's only fair you know about them:
I took stats for fighting units only. I ignored casters (HT, Ravens, etc.), suicide units (scourge, infested terrans), and casted/produced units (auto-turrets, broodlings (both games), etc.).
I ignored armor stats. This could potentially throw a huge monkey wrench in the works (if someone more dedicated feels like showing a hole, that would be great).
BW and SC2 damage systems work differently: BW has a blanket damage type vs armor type system, while SC2 (though it can be approximated on that model) is individually tuned by unit. As a "best guess" I used the maximum damage for every unit - however, I did not account for splash damage at all.
I took the average of all units' stats without regard to which units are used more or less often. Additionally, I listed all units only once, with (a) their most common use and (b) any upgrades which are gotten almost all the time. (Again, if more experienced players can correct me on these, I'd appreciate it.)
I considered cost in one set of numbers, but made no attempt to adjust for relative values of minerals and gas either within each game or between games.
The spreadsheet I will be referring to for the rest of this topic can be found here.
Summary of the Numbers
Broodwar has 28 fighting units (counting workers): 10 P, 10 T, and 8 Z. SC2 has 32: 12 P, 11 T, and 9 Z. Whether or not this is a significant increase is open to debate.
The average Broodwar unit costs 255 resources. From the Terran 208 to the Protoss 315 is a spread of 107, or 42% of the average. The average SC2 unit costs 276 resources. From the Zerg 219 to the Protoss 348 is a difference of 129, or 47% of the average. These numbers are comparable, but both the overall cost (as a number) and variance is increased in SC2.
The average Broodwar unit has 173 HP and has a dps of 12.393; the average Broodwar unit will die in 14 seconds against the average Broodwar unit; but the average Broodwar unit will spend 20 seconds killing the average Broodwar unit. (If these numbers being so different looks weird, the explanation is fairly simple: the high-damage units have much relative higher damage rates than the high-health units have higher health totals, so the dps average is "higher" than the hp average.) By comparison, the average SC2 unit has 202 HP and has a dps of 15.274; the average SC2 unit will die in 13 seconds against the average SC2 unit; again the average unit will spend 19 seconds killing the average unit.
Essentially, as much as can be determined from this approach, SC2 is comparable to Broodwar in damage and practical damage rates.
Racial Breakdown of the Fancy Numbers
The above numbers come from the HP/DPSA (hit points vs average DPS) and HPA/DPS (average hit points vs DPS) stats I put together. HP/DPSA is the hit points of a unit divided by average DPS, and indicates how long a unit will survive "on average". HPA/DPS is the reverse: average hit points divided by the particular unit's DPS. (I also averaged both of these for each race and as a total.)
Then to get an overall feel for a unit, I compared the two with an OvD - offense versus defense, or does the unit kill the average unit before it dies to it? In Broodwar, the vast majority of units have negative OvD scores - that is, they die to the average unit before they kill it. The exceptions are your usual suspects: reaver, archon, carrier, sieged tank, valkyrie (!), battlecruiser, ultralisk. The average OvD is -6.152. In SC2, this still holds true overall: average SC2 OvD of -6.094 is very comparable.
(This is where we wander into balance issues, so be warned.) However, notice that in SC2 this comes at the expense of Zerg almost entirely. Terran sits right on the average, while the average Protoss unit has a positive OvD. The full list of positive OvD for SC2: Dark Templar, Void Ray, Colossus, Archon, Immortal, Carrier, Mothership, Thor, Battlecruiser, Ultralisk.
This doesn't necessarily indicate imbalance: In Broodwar, Protoss and Terran are very comparable in OvD; Zerg is again much weaker in pure OvD. However, we know Zerg can hang in due to special abilities, lurker splash & burrow, sim city, cheaper tech/bases, and a bunch of other things I can't represent on this chart. Also I've completely ignored the air/ground distinction. However, it reinforces what we all know: Zerg is weaker on a unit-for-unit basis (in both games).
What caught my attention most, however, was the difference in HP/DPSA and HPA/DPS numbers, not as the OvD total, but individually. Broodwar Protoss are much tougher, featuring an average 18 sec HP/DPSA (defense) to the Terran 11 sec and Zerg 12 sec. However, their average HPA/DPS (offense) is 21 sec, compared to the Terran 15 sec and Zerg 24 sec.
Protoss: Strong units, a little slow about killing things (their damage-dealers are high tech). Terran: Weakest units (!) for cost, but best dps by far - marine and vulture are easily accessible as well. Zerg: Weak units as well, and honestly not that good at killing things either: forced to rely on basic unit compositions etc. as discussed above.
Again, this doesn't represent the whole picture (as above, as well as: surrounds, scouting, detection), but compare SC2:
Protoss fighting units in SC2 have both the highest average HP/DPSA (good) and lowest HPA/DPS (good) of all three races. Terran come out with the weakest units again (!), but retain a solid offensive average. Zerg units are still almost as weak as Terran, and their HPA/DPS (offense) has slipped from 24 to 25 in the new game while the other races have held even or improved.
Protoss: Strong units, now with massive damage-dealing capability. Terran: Weakest units, but still with solid DPS and (in theory) slightly easier availability of damage-dealers. I think. Zerg: Weak units, not much damage-dealing capability on a per-unit basis. Pure numbers suggest hydras ought to see more play, but I haven't accounted for speed or range in the stats I'm looking at.
Once again, this can't be construed as a balance argument - far too much information has been purposefully or necessarily left out, as noted above many times. On the other hand, I think if there were enough other evidence you could use these stats as supporting evidence that Protoss may need to be reigned in, or Zerg buffed a little. On the other hand, maybe we just need the next Boxer (or the original Boxer) to find the next dropship.
Conclusion
Things that need to be addressed before I can make solid overarching arguments from this particular numerical approach:
How to incorporate range?
How to account for the fact that some units get used more than others?
Is it even possible to account for the missing spell-casters using this methodology? In BW, for instance, storm makes up a huge part of Protoss's "missing" DPS.
However, having run these numbers, and despite my uneasiness with Protoss's lack of (numerical) weakness, I am fairly confident in saying that SC2 has mimicked Broodwar's damage/HP ratios and general damage patterns very well. I think that probably the best place for Blizzard to look for patching possibilities is in the tech tree, whether we're talking special abilities (duration/damage) or buildings (cost/build time). Some analogous changes from BW that come to mind:
The storm nerf. I'm still bitter about this one but if I'm honest after the Bisu revolution PvZ with 224 damage storm would be a bloodbath.
The 75 mineral turret. I don't know what modern TvZ would look like without this, even though I think it came before muta-stacking and was actually aimed at PvT. (Law of Unintended Consequences, much?).
siege tanks(only 100 gas) in sc1 raped dragoons 24 dps sieged vs the 17 we have now. also stalkers have better firing mechanism because of insta laser instead of the big blue balls(slow attack animation of dragoons vs stalkers) and vultures cost 75 deploying 3x 125dmg spider mines ownage. vultures were so fast and had insane dps vs workers whereas hellions are cannon fodder. you would never use zerglings against vultures because of mine drag instagib.
emp removed all archon's shield and terran also used multiple vessels(225 gas) which had so much mobility + defensive matrix. you can see why few terrans make ravens because raven + ghost is a waste of money.
new units have a lot of hp like colossus and thor. you would never use reavers vs siege tank unless you had shuttle and nowadays the insane tanking colossus can mow down siege tanks.supply wise and cost wise the stimmed marauder does more single target dps than siege tanks and that is what is more important because killing single boss units faster also meant reducing the dps you receive.
Very interesting to read. The numbers were pretty much what I expected. I don't really see SC2 units killing that much faster or slower compared to BW units. I think that banelings are a huge part of TvZ but are hard to represent using numbers because of the uncertainty of their damage.
I don't think you can ignore armor for several key reasons: armor represents a large reduction in the damage of fast-attacking, low damage-per-attack units (like zerglings and corsairs), variance in upgrades means that armor upgrades are more valuable for some units, the dual-attack nature of some units such as phoenixes and zealots, and the fact that armor does not affect Protoss shields.
Other than the armor quibble, damage modifiers against armor types, and the actual choices for armor types in SC2 are really inconsistent. It makes no sense from a flavor or gameplay sense that hydralisks are light, yet ghosts are not, and stalkers only get +4 against armored, whereas marauders get +10 (although I suppose stalkers would be way too good if it were much more than +4).
Protoss: Strong units, now with massive damage-dealing capability. Terran: Weakest units, but still with solid DPS and (in theory) slightly easier availability of damage-dealers. I think. Zerg: Weak units, not much damage-dealing capability on a per-unit basis. Pure numbers suggest hydras ought to see more play, but I haven't accounted for speed or range in the stats I'm looking at.
This summary doesn't make Z sound too good lol, where's our benefit?
That being said, I don't mean to sound mean here and appreciate the effort, but these numbers seem to me to be almost useless. Ignoring armor and abilities, range and movement speed, and troublesome units like Banelings... there's just so many critically important variables you've left out that I feel this can't possibly give us a decent representation of anything.
Nice analysis, and it's nice to see the numbers are fairly close, considering there are many units in SC2 that deal a lot of damage, like the Immortal and Colossus.
Overall, Zerg has the weakest army, but you are right in the way that you said Terran is the weakest but has the highest dps, which I'm guessing you mean weak as in HP wise.
I think casters need a look. BW had tons of potent casters. SC2 casters are neutered by comparison. Storm only does 80 damage and Khaydarin amulet still got nerfed. The ghost is about the only caster that got buffed. Ravens and infestors are absolutely terrible compared to their BW counterparts (science vessel and defiler).
You really need to consider the context of what you are saying for this to be relevant in any way. Think about the cost of these units, the way they are produced, the required tech, their role as a part of an overarching strategy, the upgrade potential and so on.
wait so protoss is actually stronger in sc2? o.O i find that incredibly hard to believe (old zlot vs new zlot, stalker vs dragoon, old ht vs new ht). though i guess the immortal and collosus skew things a tad.
I think the spellcasters are critical to the game balance. Early game toss is mainly based on sentries and using forcefields to survive. Zerg doens't even use infestors because they're so crappy and terran doesnt even need to use a raven for any matchup...
If forcefield was to receive a nerf, no one would play protoss because the gateway units are so trash and they wouldn't survive 10 minutes into the match.
On March 15 2011 14:34 Treadmill wrote: Is DPS measured against real time or blizzard time?
Edit: just that this would make a significant difference in a few of the stats, being compared between the two games.
if it was ingame time, it would just mean that all the numbers are just a bit higher for all the races. +24% to all the numbers? they'd all go up the same percentage :o
This is an interesting analysis. I think even more important than range, speed, and armor are resource cost (minerals, gas, food). People talk about "cost-efficient" units all the time. I would be interested to see which units are actually cost efficient based on their DPS and HP. I may mess around with this a little myself if I can find some time..
Ya know, I know alot of people are claiming huge imbalances exist in SC2, but I think it wouldnt take much for Blizzard to have this game balanced while still keeping it fun for the newbies (like me). But now someone needs to pull blizz heads out of their asses and tell them to stop designing this game for bronze league.
On March 15 2011 13:57 Subversion wrote: ... these numbers seem to me to be almost useless. Ignoring armor and abilities, range and movement speed, and troublesome units like Banelings... there's just so many critically important variables you've left out that I feel this can't possibly give us a decent representation of anything.
That is a kind of a problem. Consider this a first approximation - I'll be working on trying to incorporate other things as I have time.
On March 15 2011 14:16 wonderwall wrote: You really need to consider the context of what you are saying for this to be relevant in any way. Think about the cost of these units, the way they are produced, the required tech, their role as a part of an overarching strategy, the upgrade potential and so on.
There's very little way to represent that numerically, unfortunately. See above - I'm working on trying to expand this to cover things.
On March 15 2011 14:23 da_head wrote: wait so protoss is actually stronger in sc2? o.O i find that incredibly hard to believe (old zlot vs new zlot, stalker vs dragoon, old ht vs new ht). though i guess the immortal and collosus skew things a tad.
The way I measured, the fact that Protoss's new units (Immortal and Mothership especially) are high-tech high damage units is responsible for most of this. However, given that forcefield means Protoss can tech safely, this is actually fairly representative. I think.
On March 15 2011 14:34 Treadmill wrote: Is DPS measured against real time or blizzard time?
Edit: just that this would make a significant difference in a few of the stats, being compared between the two games.
I honestly don't know. BW DPS is measured in terms of damage/(cooldown frames/fps) where fps is assumed to equal 16 for a given "BW second". SC2 cooldown and dps matches liquipedia. In practical terms, however, we've found two period of time where the dps matches fairly closely; any difference is only a difference in which game plays "faster" - the games stay comparable.
On March 15 2011 14:41 Rodeo wrote: This is an interesting analysis. I think even more important than range, speed, and armor are resource cost (minerals, gas, food). People talk about "cost-efficient" units all the time. I would be interested to see which units are actually cost efficient based on their DPS and HP. I may mess around with this a little myself if I can find some time..
Based on my initial numbers, the most HP/cost and DPS/cost efficient units for each race are:
Have you thought about the average deviation? How much do each race and unit deviate from the normal values, within their own race and compared to the others. That can often be a important factor, and maybe unveil some hidden problems.
I really enjoyed reading the OP. I've been planning on doing some numbers research myself, but I've never gotten around to it. OP's analysis is quite deep and comprehensive.
We've all heard of all the more opinionated and sentimental arguments revolving around the SC2 vs BW debate. I think it's really quite interesting to see the numbers behind the debate so that we can all make more informed opinions.
Great read, just a bit of feedback; I'd love to see some graphs, charts (or at least some spreadsheets?) to help visualise the data if you do continue to do more work on this...
edit: nvm, found the spreadsheet silly me read too fast past that bit.
This is an interesting analysis, but Im not sure how relevant it is to actual gameplay. First, splash damage is the most important factor when it gets to a relatively large size army vs army fight. If say colossus were single target firing, there wouldnt even be a debate about the protoss death ball. Also, spells are too important in any large size engagement and very hard to model in that way. The facts that zerg units are very weak compared to terran units is meaningless when swarm makes them unkillable.
I can see the merits of a pure numbers approach, but I think it is too limiting to draw conclusions on the sate of balance between the two games.
Just to make the one guys reply true zerg QQ etc. etc....but being serious instead of kidding around. The numbers are interesting I can definitely give them that...also since many of the units are very similar to their BW counterparts it doesn't surprise me that on the whole the ratios etc. are very close. kind of if it aint broke why fix it mentality, yes i know the two in many ways are vastly different but a lot of the basic systems work very similarly.
While I honestly can't help you with your quest for more ways to stat compare the two it makes an interesting read.
First of all, your analysis probably miscalculated Brood Lord DPS (they spawn broodlings on attack, which do a significant deal more damage than the unit does by itself), so I would guess Brood Lord belongs on your list of positive OvDs as well. This shows us, really, that the units with high supply and/or limited mobility do good against the average unit one-on-one (which is common sense).
The problem with your averages (and why it shows toss on top) is that it doesn't take any metric into account when averaging their attributes. If the game only had 2 units in it: Immortals and Zerglings, your analysis would reveal a positive "OvD", because an Immortal beats the average between an Immortal and a zergling. However, in game, you'd find that Immortals by themselves are terrible against zerglings.
Think of it this way. You could say that in chess, there are 6 types of pieces: pawn, king, queen, knight, bishop and rook. 3 of these pieces move up to 7 squares at once, One moves exactly 3 squares, and the other 2 move 1 square at a time. You could say that the average number of squares the average piece can move is 4.3, and that bishops, rooks, and queens are therefore powerful because they move more than this, but this also neglects that half of the pieces on the board are pawns, and that the board is designed so that the pieces which can move a lot of squares almost never are able to until the game is nearly over. In short, your analysis of which pieces are valuable has no applicability to the actual game.
Statistics have metrics for a reason. If you're going to do a comparison that has any kind of statistical backing behind it, you're going to have to use them. I'd reccomend supply, personally, though mineral/gas expenditure will no doubt explain the usefulness of certain units in the early game.
Can anyone explain to me the simple difference between the BW Marine-vs-Zergling and the SC2 Marine-vs-Zergling? What I mean is in BW the zergling will kill the marine in 1v1 combat but in SC2 the marine kills the zergling. I know the marine has 5 extra HP in SC2 but it survives the fight with more than 5 HP remaining. Were their fire rates changed at all? Movement speed or ranges different? Can someone explain this to me because I'm puzzled by it.
This is a thought-provoking piece, I guess there should be a better modeling of the interactions between the units. But the mathematical model is not the easiest to get if I weren't mistaken.
If you wanted to just create some kind of basis for comparison you could choose a unit in each game, like a Zergling. And then say the unit you want to analyze with range has range X, and attack speed Y, the time it takes the Zergling to get to the unit is Z. Over X distance the unit you're analyzing does Unit Damage/Z DPS? It wouldn't be useful in a case by case but you could use it to create a chart for big picture analysis of how range plays into offense and defense.
I'm no statistician though so thats just my idea
It was an interesting read btw, I enjoyed the info about OvD's.
On March 16 2011 05:07 Ghost-z wrote: Can anyone explain to me the simple difference between the BW Marine-vs-Zergling and the SC2 Marine-vs-Zergling? What I mean is in BW the zergling will kill the marine in 1v1 combat but in SC2 the marine kills the zergling. I know the marine has 5 extra HP in SC2 but it survives the fight with more than 5 HP remaining. Were their fire rates changed at all? Movement speed or ranges different? Can someone explain this to me because I'm puzzled by it.
I believe a stimmed marine kills a ling in BW (unless stim runs out, I forget duration). An unstimmed marine has terrible DPS and dies though. The answer in SC2 - by the numbers - is that the the zergling's damage output was lowered, and while the marine's was as well it wasn't by as much and the health was increased.
On March 16 2011 04:28 Treehead wrote: First of all, your analysis probably miscalculated Brood Lord DPS (they spawn broodlings on attack, which do a significant deal more damage than the unit does by itself), so I would guess Brood Lord belongs on your list of positive OvDs as well.
Yes. This is one of the things that got ignored in my "well screw it, wtf do we do about splash?" simplification, as I noted.
On March 16 2011 04:28 Treehead wrote: The problem with your averages (and why it shows toss on top) is that it doesn't take any metric into account when averaging their attributes. If the game only had 2 units in it: Immortals and Zerglings, your analysis would reveal a positive "OvD", because an Immortal beats the average between an Immortal and a zergling. However, in game, you'd find that Immortals by themselves are terrible against zerglings.
Statistics have metrics for a reason. If you're going to do a comparison that has any kind of statistical backing behind it, you're going to have to use them. I'd reccomend supply, personally, though mineral/gas expenditure will no doubt explain the usefulness of certain units in the early game.
You'll note that on the chart I've included damages adjusted for cost and supply; however I didn't include them immediately because I'm still not sure what direction to carry this study in.
This was basically a preliminary - strip out all the complicating factors and compare what was left.
On March 16 2011 05:07 Ghost-z wrote: Can anyone explain to me the simple difference between the BW Marine-vs-Zergling and the SC2 Marine-vs-Zergling? What I mean is in BW the zergling will kill the marine in 1v1 combat but in SC2 the marine kills the zergling. I know the marine has 5 extra HP in SC2 but it survives the fight with more than 5 HP remaining. Were their fire rates changed at all? Movement speed or ranges different? Can someone explain this to me because I'm puzzled by it.
I believe a stimmed marine kills a ling in BW (unless stim runs out, I forget duration). An unstimmed marine has terrible DPS and dies though. The answer in SC2 - by the numbers - is that the the zergling's damage output was lowered, and while the marine's was as well it wasn't by as much and the health was increased.
Actually, unstimmed marine attack rate is higher in SC2 compared to BW. Stim was cut from a 100% boost to a 50% boost, though.
Zergling attack rate is much slower in SC2 compared to BW. Adrenal glands was cut from 33% to 15.7%.
On March 16 2011 05:07 Ghost-z wrote: Can anyone explain to me the simple difference between the BW Marine-vs-Zergling and the SC2 Marine-vs-Zergling? What I mean is in BW the zergling will kill the marine in 1v1 combat but in SC2 the marine kills the zergling. I know the marine has 5 extra HP in SC2 but it survives the fight with more than 5 HP remaining. Were their fire rates changed at all? Movement speed or ranges different? Can someone explain this to me because I'm puzzled by it.
The average Broodwar unit has 173 HP and has a dps of 12.393; the average Broodwar unit will die in 14 seconds against the average Broodwar unit; but the average Broodwar unit will spend 20 seconds killing the average Broodwar unit. (If these numbers being so different looks weird, the explanation is fairly simple: the high-damage units have much relative higher damage rates than the high-health units have higher health totals, so the dps average is "higher" than the hp average.) By comparison, the average SC2 unit has 202 HP and has a dps of 15.274; the average SC2 unit will die in 13 seconds against the average SC2 unit; again the average unit will spend 19 seconds killing the average unit.
So let me get this straight. The average Broodwar unit takes 14 seconds to die when fighting another of the same average Broodwar unit, and yet it takes 20 seconds to kill the average Broodwar unit.
This would seriously shit over PvP if my stalkers would kill his stalkers in 15 hits but his stalkers only needed 10 hits to to kill my stalkers. You really have to explain yourself when you use fucked up logic like this. Are you using 2 seperate units? If so, then you can't call them "THE average unit". WTF. Explain.
Marines for example, relative to other units, got waaaay stronger from the transition. Everything from a faster relative firing rate, combat shields in addition to +5 hp, reactors, medivacs (you can stick 8 rines into a dropship), and clumping make them way stronger. Rines are now like ranged cracklings on stim with more HP. Siegetanks meanwhile got weaker (though smart firing is neat).
Meanwhile, units like the muta and ling kinda got hit hard. On the one hand, you can have group 20+ mutas in to a pretty powerful ball for harass, but now without stacking (and thors anyway), they are far less effective at picking off units. Lings took everything really hard; they're virtually unchanged except everything else got stronger/faster compared to them and units now clump as well. The only better thing about them is when they can run around on creep, and they they complement banelings decently.
The average Broodwar unit has 173 HP and has a dps of 12.393; the average Broodwar unit will die in 14 seconds against the average Broodwar unit; but the average Broodwar unit will spend 20 seconds killing the average Broodwar unit. (If these numbers being so different looks weird, the explanation is fairly simple: the high-damage units have much relative higher damage rates than the high-health units have higher health totals, so the dps average is "higher" than the hp average.) By comparison, the average SC2 unit has 202 HP and has a dps of 15.274; the average SC2 unit will die in 13 seconds against the average SC2 unit; again the average unit will spend 19 seconds killing the average unit.
So let me get this straight. The average Broodwar unit takes 14 seconds to die when fighting another of the same average Broodwar unit, and yet it takes 20 seconds to kill the average Broodwar unit.
This would seriously shit over PvP if my stalkers would kill his stalkers in 15 hits but his stalkers only needed 10 hits to to kill my stalkers. You really have to explain yourself when you use fucked up logic like this. Are you using 2 seperate units? If so, then you can't call them "THE average unit". WTF. Explain.
averages get skewed by high damage/high health units.
The average Broodwar unit has 173 HP and has a dps of 12.393; the average Broodwar unit will die in 14 seconds against the average Broodwar unit; but the average Broodwar unit will spend 20 seconds killing the average Broodwar unit. (If these numbers being so different looks weird, the explanation is fairly simple: the high-damage units have much relative higher damage rates than the high-health units have higher health totals, so the dps average is "higher" than the hp average.) By comparison, the average SC2 unit has 202 HP and has a dps of 15.274; the average SC2 unit will die in 13 seconds against the average SC2 unit; again the average unit will spend 19 seconds killing the average unit.
So let me get this straight. The average Broodwar unit takes 14 seconds to die when fighting another of the same average Broodwar unit, and yet it takes 20 seconds to kill the average Broodwar unit.
This would seriously shit over PvP if my stalkers would kill his stalkers in 15 hits but his stalkers only needed 10 hits to to kill my stalkers. You really have to explain yourself when you use fucked up logic like this. Are you using 2 seperate units? If so, then you can't call them "THE average unit". WTF. Explain.
averages get skewed by high damage/high health units.
This is still ambiguous.
Does it mean "The average expected life span of a randomly chosen unit taking damage from another randomly chosen unit is 14 seconds. The expected time required for a randomly selected unit to kill another randomly selected unit is 20 seconds"???
Really cool to see the numbers on this. Super interesting to see the changes in units from BW to SC2.
Dunno what it all means though - you say Protoss has the best units (DPS and HP for money wise), if this is the case then why does Protoss rely so heavily on spellcasters? With Force Fields and Storms, it seems a pretty strong counterargument that Protoss units aren't as strong as they seem, since without spells they lose most head-on engagements.
In SC2, micro is more prevelant, and this nullifiies a lot of your argument IMO. For instance, a Zealot is wins cost for cost against a Marauder, yet will lose with kite micro. Kiting in SC2 is soooo important, yet rarely happened as extensively in BW. Micro increases the effectiveness of your units, and some units are easier to micro than others (i.e. kiting with a Sentry rather than a Marine)
On March 16 2011 13:28 Barca wrote: In SC2, micro is more prevelant
Are you sure about this?
I agree with the argument that micro makes numbers less important, since micro is different in each game. But on what basis is micro in SC2 more prevalent?
Please don't let this derail, PM me if you have an argument.
While I started by comparing simple DPS average, this seemed oversimplified. The measure I eventually went with was DPS compared to the average hit points of all units in the study. This has its limitations and inaccuracies, of course, but it seemed to give the most reasonable numbers for comparison within each game - further thoughts will be detailed later.
I performed a similar exercise recently, but used slightly different methodology, and I only used SC2 data. My purpose was to figure out if certain units were just being overhyped (after all the cries of 'marauders are OP' and 'nerf void rays' that seemed to be everywhere at one point), but I'll present it here because it could easily be used to compare SC2 units to BW units if people think it's a worthwhile metric.
What I did was take certain stats from a unit:
Mineral Cost Gas Cost Food Cost Health Shields Damage Per Attack Number of Attacks Attack Cooldown
I then subjected each unit to an arbitrary amount of damage against them (10dps), and worked out how long each unit would survive against that dps, and how much damage that unit would do before death. I called this 'base damage before death'. From that figure, I then calculated base damage done per 100 minerals, 100 gass], and 1 food for each unit, calling them mineral, gas, and food efficiency indices (since the numbers themselves are meaningless, but the numbers relative to other units are useful).
Anyway, an example of what I did:
Marauder (Stimmed)
Mineral Cost - 100 Gas Cost - 25 Food Cost - 2 Health - 125 Shields - 0 Damage Per Attack - 10 Number of Attacks - 1 Attack Cooldown - 1s
Base DPS - 10 Survival Time Against 10dps - 12.5s Base Damage Done Before Death - 125
Mineral Efficiency Index - 125 Gas Efficiency index - 500 Food Efficiency Index - 62.5
A couple of obvious flaws are that I haven't taken splash damage or casters into account (so the unsieged tank scores more highly than the sieged tank, hellions seem awful when they can actually roast an entire mineral line in seconds, and HTs don't feature at all), and I haven't yet bothered to work out carrier/interceptor damge (brood lord/broodling was relatively easy, so I imagine carriers will be also). But overall, I think it's a decent enough way of comparing how most units will cope with an old fashioned stand-up brawl.
It's unsurprising that T3 units rule the roost in terms of all three efficiency indices, simply because they have so much health and dps (as long as you can actually get them to engage, which can be a particular issue for thors and ultras because of their size).
Now, these lists are for base dps only - I do have the data for how units perform against various armor types but I've been typing and editing this for ages already and I want to stop . Either way, I think it's still a worthwhile way of comparing the relative general strengths of certain units. I certainly hadn't realised quite how resource-efficient banshees were, even against non-light units, nor did I realise just how important it is to keep your mutalisks away from the enemy army - they're a truly awful use of resources in a 'proper' fight.
tl;dr
Slightly different way of comparing units - if you find it useful, then great. If not, then at least this was a fun way of bosting my post count by 1. WIN-WIN :D
On March 16 2011 11:14 Geovu wrote: This would seriously shit over PvP if my stalkers would kill his stalkers in 15 hits but his stalkers only needed 10 hits to to kill my stalkers. You really have to explain yourself when you use fucked up logic like this. Are you using 2 seperate units? If so, then you can't call them "THE average unit". WTF. Explain.
First another example, then the math (I'll spoiler it), and then a discussion of what I may have done wrong.
Take two random sets of numbers we call A and B. We'll use A = {2, 3, 4} and B = {8, 9, 10} to make things simple. The mean (average) of A is 3; the mean of B is 9. Mean(A)/Mean(B) is 1/3 (and Mean(B)/Mean(A) is 3. If instead we take (2/9+3/9+4/9) = 9/9 = 1, then divide by 3, we still get 1/3; but if we take 3/8+3/9+3/10 = (135+120+108)/360 = 363/360. So when we divide by 3 this time, we get 121/360, a little more than 1/3.
In other words, if your two averages are different to begin with, taking them "apart" and comparing them individually is going to get you weird numbers.
Take two sets, each with n members. Let A = {a1, a2, a3, ..., an} and B = {b1, b2, b3, ... bn}. Then mean(A) = sum(A)/n, and mean(B) = sum(B)/n.
The ratio of the means will be mean(A)/mean(B), also equal to sum(A)/sum(B) since both sets have the same number of members n.
Any given member of a set has a ratio of any member of a set to the mean of the other set as well, for example Ax/mean(B). The mean of these will be [A1/mean(B) + A2/mean(B) + A3/mean(B) + ... + An/mean(B)]/n = [(A1 + A2 + A3 + ... + An)/mean(B)]/n. But the sum of all terms of A is sum(A), mean(B) is sum(B)/n, so this is nsum(A)/nsum(B) = sum(A)/sum(B) = the mean.
[This is the HP/DPSA stat - HP divided by the average DPS. So the average will equal the average HP divided by average DPS.]
However, we can also take the ratio of the mean of a set to any member of the other set, for example mean(A)/Bx. If we find all of these ratios and take the average we'll have: [mean(A)/B1 + mean(A)/B2 + mean(A)/B3 + ... + mean(A)/Bn]/n. But this sum is harder to find since the denominators aren't already equal. To get the sum, for each term we need to find mean(A)*[product(B)/Bx)]. So the entire thing will be ({mean(A)*[product(B)/B1 + product(B)/B2 + product(B)/B3 + ... + product(B)/Bn]}/{product(B)})/n. Remembering that mean(A) = sum(A)/n, we find ({sum(A)*[product(B)/B1 + product(B)/B2 + product(B)/B3 + ... + product(B)/Bn]}/product(B))/n^2, which in all probability will not equal sum(A)/sum(B).
[This is the HPA/DPS stat - average HP divided by DPS. The average HPA/DPS will often be different than the average HP/average DPS.
What it means: note that the second average can be bigger or smaller (even if it's usually bigger in this comparison. What I *think* this indicates is something about the variance in stats. I postulate - can't prove - that if the average is heavily skewed upwards by one or two units, the second measure will be higher than the first. (2/11 = 0.18; but (2/5 + 2/3 + 2/25)/3 = (86/75)/3 = 0.38. If the average is skewed downwards, it should be lower than the first. In other words, OvD - the comparison - can indicate the overall skew of the DPS distribution.... or something.
I'm not sure neglecting armor/size (at least in BW) is wise. The damage variance they create is large enough that I'm not sure you can make these sort of average statements. Anything with concussive damage (Vultures, Firebats, Ghosts) often has quartered or halved DPS. Your data may be slightly more accurate if you took the percentage of units in BW with each "weight class" then weighted the damage accordingly; this should be relatively easy to do. Capturing splash damage (which makes the Valkyrie, Firebat, and Corsair not totally crappy despite low DPS) is another issue I can't think of an easy way to account for.
While I started by comparing simple DPS average, this seemed oversimplified. The measure I eventually went with was DPS compared to the average hit points of all units in the study. This has its limitations and inaccuracies, of course, but it seemed to give the most reasonable numbers for comparison within each game - further thoughts will be detailed later.
I performed a similar exercise recently, but used slightly different methodology, and I only used SC2 data. My purpose was to figure out if certain units were just being overhyped (after all the cries of 'marauders are OP' and 'nerf void rays' that seemed to be everywhere at one point), but I'll present it here because it could easily be used to compare SC2 units to BW units if people think it's a worthwhile metric.
What I did was take certain stats from a unit:
Mineral Cost Gas Cost Food Cost Health Shields Damage Per Attack Number of Attacks Attack Cooldown
I then subjected each unit to an arbitrary amount of damage against them (10dps), and worked out how long each unit would survive against that dps, and how much damage that unit would do before death. I called this 'base damage before death'. From that figure, I then calculated base damage done per 100 minerals, 100 gass], and 1 food for each unit, calling them mineral, gas, and food efficiency indices (since the numbers themselves are meaningless, but the numbers relative to other units are useful).
Anyway, an example of what I did:
Marauder (Stimmed)
Mineral Cost - 100 Gas Cost - 25 Food Cost - 2 Health - 125 Shields - 0 Damage Per Attack - 10 Number of Attacks - 1 Attack Cooldown - 1s
Base DPS - 10 Survival Time Against 10dps - 12.5s Base Damage Done Before Death - 125
Mineral Efficiency Index - 125 Gas Efficiency index - 500 Food Efficiency Index - 62.5
A couple of obvious flaws are that I haven't taken splash damage or casters into account (so the unsieged tank scores more highly than the sieged tank, hellions seem awful when they can actually roast an entire mineral line in seconds, and HTs don't feature at all), and I haven't yet bothered to work out carrier/interceptor damge (brood lord/broodling was relatively easy, so I imagine carriers will be also). But overall, I think it's a decent enough way of comparing how most units will cope with an old fashioned stand-up brawl.
It's unsurprising that T3 units rule the roost in terms of all three efficiency indices, simply because they have so much health and dps (as long as you can actually get them to engage, which can be a particular issue for thors and ultras because of their size).
Now, these lists are for base dps only - I do have the data for how units perform against various armor types but I've been typing and editing this for ages already and I want to stop . Either way, I think it's still a worthwhile way of comparing the relative general strengths of certain units. I certainly hadn't realised quite how resource-efficient banshees were, even against non-light units, nor did I realise just how important it is to keep your mutalisks away from the enemy army - they're a truly awful use of resources in a 'proper' fight.
tl;dr
Slightly different way of comparing units - if you find it useful, then great. If not, then at least this was a fun way of bosting my post count by 1. WIN-WIN :D
Honestly this probably is worthy of an OP. I have only a couple criticisms. First, you're overvaluing the melee units (because they can be kited) and second, by listing the mineral and gas costs separately, you overvalue those whose cost is tied up in both.
I looked over the spread sheet and I'm pretty impressed with the amount of data you've managed to gather so far. However, for anyone to be able to draw any sort of useful conclusions there are just so many variables left out when you look at the game purely numerically and disregard all the special traits and characteristics of each race.
Stuff like protoss rushing to 2-2 in a very short time interval due to CB or various aoe damages and special abilities etc etc. I'm just not sure exactly what you're trying to accomplish.
The game obviously isn't meant to have units of each race stack up equally in cost effectiveness, damage and so forth (as I'm sure you're aware). The macro mechanics and the way the macro of each race is designed can be viewed as a form of equalizer that makes up for the short-comings of the race by extending it in other areas in terms of "expandability", econ, production blah blah.
I like the information presented here, but I'm not sure exactly how it is applicable in it's current form. If I were you, I'd take the numbers you've found and divide them among average unit costs and/or build times to maybe better paint a picture of where the races stand in relation to each other. I suspect that zerg are being less screwed over if these things are taken into consideration.
Also, specific unit costs should be altered to include the costs of developing that unit, which I think will help better account for the macro variables.
For instance, an HT costs way more than it's base price of 50/150, considering you also need to build pylon -100, gateway - 150, twilight - 150/100, archives - 150/200 and then research storm - 200/200... meaning 800/650 to get 1 storm-capable templar on the field (1450 total resources, all things being even).
Compared to say Infestor which uses an overlord - 100, spawn pool - 200, lair tech - 150/100, infestor pit - 100/100, then infestor 100/150... total cost 650/350 to get an infestor on the field (1000 total resources, all things being even).
I don't know, just maybe some insight on what could be done with this info...
It's interesting statistics, but not really applicable for any balance discussion. Splash, bonus damage, range, spells and passive abilites all are important factors to consider. E.g., DPS from a marauder would differ against melee units depending on if concussive shells were researched or not. I don't know how such stats could be applied though, probably too hard to make an overall analysis like in the OP, maybe for individual units, but the work would be a bit too much I guess.
Interesting how costs have increased since BW though, but I think mining is more efficient now and might make up for it?
The problem with this is that you tend to average many units. Protoss units definitely feel weak-ass in SC2, simply because Terrans and Zerg now have units that have comparable hp.
Thus, let's compare the "bread and butter units".
Request stats of "bi modal" units of equal supply: Marine + Maurauder vs. Roach + Hydra vs. Zealot Stalker. For kicks, let's do 40 supply of each race, and 20 supply of each unit.
Then at the end display the stats as well as the resources consumed.
My Hypothesis: It's going to come out with the results of what most of us feel when playing SC2. MMM is much more crazier than in SC1.
On March 16 2011 05:07 Ghost-z wrote: Can anyone explain to me the simple difference between the BW Marine-vs-Zergling and the SC2 Marine-vs-Zergling? What I mean is in BW the zergling will kill the marine in 1v1 combat but in SC2 the marine kills the zergling. I know the marine has 5 extra HP in SC2 but it survives the fight with more than 5 HP remaining. Were their fire rates changed at all? Movement speed or ranges different? Can someone explain this to me because I'm puzzled by it.
Unstimmed = same Stimmed BW = 100% increase ( 40 hp all game ) Stimmed SC2 = 50% increase ( up to 55 hp ) Lings Normal / Adrenaline BW = 3 and 4 hits/sec or 8 / 24 frames and 6/ 24 frames ( " Fastest " ) Lings in SC2 hit for 1 / 0.70 and 1 / 0.56 and then multiplied by gamespeed SC2 lings come with auto surround ( Jaedong/July don't really need this though ) Speed on lings in SC2 have a + ~0.20 http://wiki.teamliquid.net/starcraft/Unit_Movement_Speed ( bw stats ) all based on relations to workers
On March 16 2011 05:07 Ghost-z wrote: Can anyone explain to me the simple difference between the BW Marine-vs-Zergling and the SC2 Marine-vs-Zergling? What I mean is in BW the zergling will kill the marine in 1v1 combat but in SC2 the marine kills the zergling. I know the marine has 5 extra HP in SC2 but it survives the fight with more than 5 HP remaining. Were their fire rates changed at all? Movement speed or ranges different? Can someone explain this to me because I'm puzzled by it.
Unstimmed = same Stimmed BW = 100% increase ( 40 hp all game ) Stimmed SC2 = 50% increase ( up to 55 hp ) Lings Normal / Adrenaline BW = 3 and 4 hits/sec or 8 / 24 frames and 6/ 24 frames ( " Fastest " ) Lings in SC2 hit for 1 / 0.70 and 1 / 0.56 and then multiplied by gamespeed SC2 lings come with auto surround ( Jaedong/July don't really need this though ) Speed on lings in SC2 have a + ~0.20 http://wiki.teamliquid.net/starcraft/Unit_Movement_Speed ( bw stats ) all based on relations to workers
Marines fire a lot faster in SC2 than in BW, makes a huge difference when it comes to range vs melee. Also marines in SC2 can stutter step, BW marines can't (as well).
EPT
silly me read too fast past that bit.
Thor (Ground) - 625
Brood Lord - 552
Archon (HT) - 513
