LOL Not to troll or anything, but with that analysis and unit comprehension i'm surprised that your gold. What race do you play, whatever it is, it's OP. You made horrible accusations on all the races.. .and only legit thing i saw what that Mutas are a bit pricey and they are in SC2... Every unit has its use in its given time...
I appreciate the effort, but this isnt worth the effor to pursue... There are mathmatical calculations for a lot of this stuff, check out this series...
UC Berkley course for Starcraft.. Check out 5:27, starts to mention the effectiveness mineral wise for Archons vs Zerglings.
Just a couple things i would like to point out:
IMO, VR are the only thing which doesn't have much of a weak point... Im not sure you calculated this, but the charged damage on the Void Ray... I have watched TONS of matches, and you dont see them getting charged up much... because in mass, they tend to kill stuff to fast to charge, but when they charge up, they are unbeatable...
Not know the uses of a Hellion, on top of a simple 1 to 1 ratio for everything is pointless.
SCV vs Probe. Scv 5 damage 45 HP.... Probe 5 damage 20hp 20 shield... SCV ALWAYS WINS!!!!!.... not really... if you attack 4x SCV is at 25hp and probe at 20/0... you pull the probe back for 10 seconds, and then SCV is at 25hp and Probe is at 20/20... All in all units with shield / regen have serious advantages for harassment.
Didnt take into for Possible Advantage of stealth... DT rock face when there is no detection... and actually less effective then a legs upgraded zelot... Burrowed roaches TEAR UP TANKS... but without burrow... they cant even hit them...
I was watching someones stream which was a TvT ... and his opponet went mass BC, something like 8 BC's on the attack... the player was worried for a second...but ended up making a bunch of Vikings he probably had around 14 of them... I cant quite remember the cost of everything, but i believe the BC was worth a lot more... well... the Vikings have 2 more range... he was able to kill all 8 bc's wihout losing a single Viking... With proper micro... Without the micro ... i bet the BC's would have won... I won a game going mass BC and beat Vikings because i would go in and kill a couple and repair. They never got a critical mass to 1 or 2 shot my BC's so they would go in and a couple would die and they would never kill any of mine, unless they are willing to lose 4-5.
On July 24 2010 15:58 Zev wrote: You made horrible accusations on all the races.. .and only legit thing i saw what that Mutas are a bit pricey and they are in SC2... Every unit has its use in its given time...
Really? Please point out which ones specifically and how. I'll update accordingly. Thank you so much for your insight.
On July 24 2010 15:58 Zev wrote: There are mathmatical calculations for a lot of this stuff, check out this series... [...] its the UC Berkley course for Starcraft..
Right you are. I remember when Sirlin posted reviews of each class. I especially found the flux statistic interesting.
On July 24 2010 15:58 Zev wrote: Im not sure you calculated this, but the charged damage on the Void Ray [...] but when they charge up, they are unbeatable...
Yeah, and yes I did calculate that.
On July 24 2010 15:58 Zev wrote: SCV vs Probe. Scv 5 damage 45 HP.... Probe 5 damage 20hp 20 shield... SCV ALWAYS WINS!!!!!.... not really... if you attack 4x SCV is at 25hp and probe at 20/0... you pull the probe back for 10 seconds, and then SCV is at 25hp and Probe is at 20/20... All in all units with shield / regen have serious advantages for harassment.
I accommodated this in my original post by saying Protoss is actually stronger than the data suggests.
On July 24 2010 15:58 Zev wrote: Didnt take into for Possible Advantage of stealth... DT rock face when there is no detection... and actually less effective then a legs upgraded zelot...
Wow a stalker has the same DPS as a marine.. that is so retarded and why its so hard to counter battlecruisers and air in general with them. Unstimmed even... atleast i learned that if im looking for damage dealers sentries almost equal stalkers
The most cost effective unit is the Missile Turrent and you didn't even include it in your tables. You also haven't taken into account vital upgrades/abilities, like stim pack.
Carriers are 450/250 total, 16 attacks by 5 damage every 3 seconds, so: Resources: 700; DPS = 16*5/3 = 26.(6); HP+Shields = 450 DPS*(HP+Shields)/Resources = 17
(DPS*H)/R = ((Damage/cooldown)*(HP+shield))/(minerals+gas), or cost-effectiveness: DPS over the course of a battling unit's life per dollar. I multiply here because every unit of life means another unit of time that a unit can actuate DPS. This is the meat and gravy. It thoroughly reconciles the costs (resources) with the benefits (damage, cooldown, and health).
Sorry, but this formula looks like nonsense to me. If you put in the value of two units the cost effectiveness suddenly doubles. This is because the strength of units grows quadratically with their number. So your formula puts expensive units as better than they are.
Use Sqrt(DPS*H)/R or DPS*H/R^2 instead. This formula is still not perfect, but at least it works at all.
Honestly, the people who "Noted hidden costs" on the Terran units are obviously not honest with themselves.
Terran units are easily the most cost effective units.
How could you tell? Just look at Thor and Siege tanks kill counts. Bust out some replays if you have to. They are far often WAY higher than all the other units. Anyone who says otherwise is just lying to themselves.
Even Marines - there are no starter-units for any of the other races that have the insane upgrades Marines do, and are so effective once reaching critical mass. Due to their small size and stim they easily do the highest focus-fire damage in game once reaching critical mass. Even charge Zealots cant compare due to their melee only, and Lings are far weaker in SC2 late game than they were in SC1. What could show a race to be more cost-effective than their basic 1-population 50mineral no gas unit?
Not to mention that Terran armies are usually much less gas-heavy for the damage they do. Look at the army resource counters in replays and you will find they destroy armies of far higher resource values, due to the lower gas costs.
On July 24 2010 19:00 Cerion wrote: Marines and marauders being in the bottom third of that list should really be a hint that the model oversimplifies combat.
Agreed. The chart does not account for things such as critical mass, focus firing, smart-aiming, and micro capabilities. So it's nowhere near accurate.
But as mentioned earlier, just open up some replays and look at the units kill counters, and it's simple to see who has the most efficient units.
This info is not very useful in terms of actual in game unit utility. This reminds of of bogus martial arts schools which teach you how to defend against a punch in some fancy way when you know exactly how its coming. The students then get their asses beat in a street fight and wonder what happened. The fact is, the model you are using is inaccurate compared to how actual games play out due to multiple variables that you are not accounting for (just like in the martial arts example).
Using DPS*HP/Cost is clearly biased against low cost units, so as MasterOfChaos suggested here's a spreadsheet with DPS*HP/Cost^2 used: link +Upg includes the bonus damage and maxed attack upgrades against a target with 3 armor.
Far more meaningful than DPS is the number of whole hits it takes to kill a particular unit and the time that takes. That would take a much bigger table though.
Oh, and certain results, like the Ultralisk and Zergling, don't take into account how so much of their HP is spent as they actually try to reach their targets.
Still, with this there's interesting results to be found.
Unfortunately for this analysis, Starcraft 2 is not a game where each player moves a single unit to the centre of the map, force-moves them up to melee range, and then attacks; then the loser replaces their solitary unit, they move to melee range again, and the fight continues... until one of them is mined out.
This analysis is almost entirely useless because it ignores spellcasting, range, splash damage, ground/air, armour and especially mobility. Pick any unit from near the top and any unit from near the bottom and some or all of these factors will explain why they're equivalent.
Mobility is the biggest one. It allows you to scout with combat units; retreat from losing battles; kite; choose your battles; abuse cliffs; get good arcs; do run-bys; hit mineral lines; harass bases; and generally gain map control. Note how lots of the units at the bottom of the lists (like sentries, stalkers, phoenixes, mutalisks, corruptors, hydralisks, hellions, reapers, marines) are extremely mobile, or in the case of the sentry allow you to control your enemy's mobility. It's that important.
The one interesting thing you've shown is how strong battlecruisers are if you can reach them. They rate highly on damage; they can abuse air mobility against a ground-heavy opponent; they're ranged; they have 3 armour even before upgrades; they can shoot air and ground; and they have a decent damage spell. However, note that rating highly on damage is only one of those factors, and might not be the most important. It's the whole package that matters.
Here's an example: Zealots alone can do very little against stalkers, phoenixes, void rays, colossi, dark templars, archons, motherships, mutalisks, corruptors, hydralisks on creep, zerglings, roaches, ultralisks, hellions, reapers, vikings, ghosts, marines, marauders with concussive shells, sieged siege tanks, and banshees. Those are all units they're more 'cost effective' than. That's 21 different units - almost all of the combat units in the game! - which your analysis is able to be hopelessly wrong for. They also lose to banelings, which don't fit into your model at all. These aren't minor issues! They utterly dominate the thing you're trying to measure. This is why your analysis is unhelpful, and its focus on DPS*H/R may even give new players the wrong idea and screw up their play until they learn better.
Except... even looking at the whole package isn't enough. Take marauders in PvT, for example. Marauders beat stalkers on your DPS*H/R measure; they beat zealots on mobility, which you ignore. But... a 50:50 mix of zealots and stalkers beats marauders.
A beats B, A beats C, but A loses to half of each. That kind of situation makes a mockery of the fundamental idea of singular unit comparisons.
It's such a nice idea in theory until you think, OBVIOUSLY the more expensive units do more damage, take more time to build. OBVIOUSLY 1 gas is more valuable than 1 mineral since you can have 24 workers mining 8 patches of minersals but only 6 workers getting you gas. In the end none of this is useful in practice since you wont use ANY of this data during a game, you still go by the situation and adapt accordingly.
On July 24 2010 17:08 MasterOfChaos wrote: Use Sqrt(DPS*H)/R or DPS*H/R^2 instead. This formula is still not perfect, but at least it works at all.
Ah, I see. I'm effectively squaring the first half, so I need to accommodate for that. Thank you for your help! I'll fix this.
On July 24 2010 22:14 randomnine wrote: This is why your analysis is unhelpful, and its focus on DPS*H/R may even give new players the wrong idea and screw up their play until they learn better.
Oh gosh I hope not, since that certainly wasn't my intention. Hopefully new players will have read the disclaimer which explains why not to take it at face value. One must bear in mind other factors and judge units accordingly, as you mention. But if people really think simply having objective data posted on Teamliquid hurts the community more than help, I won't hesitate to delete this post. It was more an exercise for my own better understanding of the game, anyway.
On July 24 2010 22:37 carwashguy wrote: One must bear in mind other factors and judge units accordingly, as you mention. But if people really think simply having objective data posted on Teamliquid hurts the community more than help, I won't hesitate to delete this post. It was more an exercise for my own better understanding of the game, anyway.
Sorry. I came on a bit strong.
I think this kind of approach is helpful, but that you're being a bit too reductionist. Weighting gas against minerals, for example? There is no correct weighting. Availability of gas and minerals fluctuates through a game, and there's no exchange rate for turning one into the other. You simply have to stop using it when it runs out. Even ignoring fluctuation in resources, the value of gas depends entirely on how badly you want or need specific gas-heavy units or tech; the value of minerals depends on how hard you want to power and add basic production buildings.
Great catch by MasterOfChaos on the quadratic scaling of units in numbers, actually. I admit, I'm curious how that'll shake things up.
There's not much 'analysis' in the OP at all. It's basically, cheap units are good and expensive ones are good with good micro or in correct conditions.
The only good way of doing a cost-effectiveness analysis is by using armies vs armies. Versatility is the key, not mass 1/2 unit types and die Idra-style.
EPT
