|
United Kingdom13775 Posts
It's not controversial to say that Terran has been the historically dominant race in BW. Indeed, Terran enjoys the most favorable non-mirror winrates, the most MSL AND OSL titles, and the most bonjwas of any race by far. Is this a result of the innovation of a few players, and something that could have gone a whole different way if, say, Boxer were a Protoss genius rather than a Terran one? Or are players mostly a product of their environment, including their race and the maps?
![[image loading]](http://i.imgur.com/WciMZ1I.jpg) Unstoppable war machine?
Balance and Bonjwas
Four years ago, I made the Balance and Bonjwas article, an attempt to answer this question. There were a few general conclusions that that article came to. In general, the conclusion was that the cause of Terran dominance could be attributed to a few simple factors:
1. The TvT is the most "skilled" mirror matchup, so the most skilled Terrans have a higher chance of rising in tournament ranks.
2. The "weak" matchup, PvT, is a much weaker race-based advantage than TvZ or ZvP.
3. There are more Zerg than Protoss players, so Terrans get a better chance to minimize their weaknesses.
4. Many small factors compound within tournaments, including those mentioned in (1), (2), and (3), to give top Terrans a substantial advantage in winning titles and attaining the rank of "bonjwa."
That article, while bringing up a lot of interesting points, was ultimately too simple to be able to fully consider all of the issues related to balance within BW. So this sequel is part of an attempt to look into the balance issue in much more depth. One article is too short to properly address this issue, and this sequel will be completed in multiple parts.
Before I continue, I want to note that in the Brood War scene, it generally isn't considered popular to talk about the issue of balance. The idea is that the balance in the game is so close to perfect that skill matters more than the race you choose. While I sympathize with that viewpoint, I will simply have to say that it is quite a bold claim to make that three races as asymmetric as BW's Terran, Zerg, and Protoss, could all be perfectly balanced. In fact, this does seem to be contrary to reality, and while skill is very important, so is the set of tools that your race gives you to work with. So, in other words, race matters, and therefore so does balance.
Maps and Metagames
This article is going to focus on two major issues: map balance and ELO rankings. In a more fundamental sense, these two issues are a way to analyze how "skilled" players are, and how they are influenced by the environment they are in. We will look at how skewed maps tend to be (i.e. how much players tend to be influenced by favorable and unfavorable maps), winrates on individual maps, and winrates in matchups over time. We will also look at trends in ELO peaks between races for overall and matchup-specific ELOs to determine which matchups are more vs. less favorable, and more skill-based vs. more luck-based. The results will be summarized in the conclusion, which is what I'd recommend you skip to if you don't have the time to read the entire article.
Technical Notes
As with the Balance and Bonjwas article, this one will involve statistical analysis. However, for those who are not strongly educated in statistics, this snippet explaining the nature of the normal distribution should be sufficient background to be able to follow along.
The data set considered for this study will be all games played within the KeSPA scene from May 18, 2001 (the release of patch 1.08) to September 22, 2012 (the last of the KeSPA BW scene). I don't think that the 2012 cutoff will generate much controversy. The Patch 1.08 cutoff, however, was somewhat contentious within the previous thread, with the argument being that certain mechanics (including the invention of mutalisk stacking) were so game-changing that they redefined the game and made previous balance concerns obsolete. However, as pointed out by Mortality, the issue with this line of thinking is that all such metagame changes were very much gradual, being developed and refined by many players over the course of multiple years. And as such, the only logical cutoff point is an event that is both instantaneous and universal: Patch 1.08.
All of the data for this study was taken manually from the BW Korea TLPD archive. One slight issue with this is that there was no easy way to organize the TLPD data by year; only by map. As such, the "year" associated with any given game is the release year of the map on which it was played. Given that most maps lasted only a few months, this is generally accurate. However, simply note that the years are not exact, so beware of making any statements about the timeframe of any given event; only about general trends. For example, while this data is precise enough to say "trend X is associated with the swarm season" it would be much less reasonable to assert that "trend Y is associated with the OSL final between Boxer and Anytime." The data simply is not precise enough to be able to associate information with so specific an event.
There are a grand total of 38 graphs in this OP. For the sake of reducing the size of an already rather large post, I'm going to spoiler all of them.
Map Data
The following graph shows the distribution of maps by maximum player count:
+ Show Spoiler +
Perhaps unsurprisingly, 4-player maps are the most popular, followed by 2-player then 3-player maps. The other sizes for maps are mostly unused.
How is the balance in this map? The next three graphs depict the balance of non-mirror matchups, organized by the three "favored" MUs (TvZ, ZvP, and PvT):
+ Show Spoiler +
(Note: maps with 10 or less games in any MU were omitted because they can very easily be skewed simply by virtue of not having many games on them, and because those tend to be the maps that are so imbalanced that they are quickly taken out of the map pool).
The averages for the three matchups are about what you would expect - slightly over 50% for each. ZvP appears to have the most favorable conditions on average (about 56% average winrate by map), followed by TvZ (~55%) and PvT (~52%). Incidentally, this is roughly in line with the performance of each race by matchup in terms of games played over all maps.
It is also interesting to look at the spread in results by each matchup, which could help to understand how maps tend to vary for each matchup. The TvZ matchup appears to be the closest to a normal spread in results, with a very high amount of games in the 50-60% range but a pretty decent spread in either direction for further deviations from the mean winrate. For ZvP, most maps are solidly in the 45-60% range, with very few below 45% and a substantial number above 60%, suggesting that strongly Z-favored maps are rather commonplace in this matchup. PvT has a big concentration, also in the 45-60% range, but with quite a few more skewed maps in both directions (but surprisingly, no maps with >85% winrate that met the cutoff).
What if we look at just the most skewed matchup - both the "favorable" and "unfavorable" ones - to see how "balanced" maps tend to be?
+ Show Spoiler +
The first criteria leads to a ~68% average, while the second criteria leads to a ~73% average. That's a fair bit more than I would have expected - it seems quite a few maps, even ones that see play, tend to have at least one rather dramatic imbalance. It surprised me further that the second criteria led to a higher average, since my expectation would be that if one matchup were so favorable, that players would shy away from playing that map if it didn't favor them, so there would be a lack of players for the most skewed MU. The biggest cluster of maps seems to be in the 60-70% range on both graphs, with quite a few in the 70-80% range and the 50-60% range.
Trivia Quiz:
There is one map in this graph with less than 52% in its most favorable matchup - which map is it? + Show Spoiler +Arcadia 2 (51.92%), with many fan favorites - Tau Cross (52.03%), Circuit Breaker (52.99%). Blue Storm (53.15%), Neo Aztec (53.33%), and Fighting Spirit (53.39%) - barely not making the cut.
There is also one map that didn't make the cut, but has the honor of being the only BW map with perfect balance - which map is it? + Show Spoiler +
Game Data
Next, we're going to look at all of the games played over the course of BW's history. As previously mentioned, the year corresponds to the release year of the map that the game was played on. Which is mostly in line with the actual year, since few maps survive more than one year in rotation, but also not 100% accurate.
First, let's look at the total games played, organized by mirror vs. non-mirror MUs.
+ Show Spoiler +
Non-mirror MUs are by far more popular, with about 3/4 of all games played. All in all, about 40k total games, 10k of which are mirror MUs. The most active years of BW seem to be 2006-2009, around the "golden age" years of BW, with a precipitous decline after 2010 leading into the waning years of the BW proscene.
Next, we look at just non-mirror games by race:
+ Show Spoiler +
So TvZ is the most popular non-mirror, with ZvP and PvT being about equally popular. This trend seems to be pretty steady throughout history. Possibly related to the idea that "the meta is defined by the TvZ."
Then, mirror games:
+ Show Spoiler +
For reference, TLPD has 87 Terran, 108 Zerg, and 70 Protoss players in its ELO list. This is about what you would expect, though perhaps it is interesting that in the 2006-2008 era, TvT was more popular than ZvZ. If you look at it cumulatively, ZvZ > TvT > PvP in terms of games played, and that shouldn't be a surprise.
Finally, let's look at winrates.
+ Show Spoiler +
On the aggregate, we get the favorable matchups with ZvP (54.7%) > TvZ (54.4%) > PvT (52.6%). Interestingly, ZvP has become a much more favorable matchup over the years, while TvZ has had a slight, but noticeable, decline in its winrate. If you look at individual years, you will see a much more erratic trend, where matchups tend to flip in favorability rather often. It might be interesting to try to match those trends to the "eras" of BW to see if they roughly coincide.
ELO Data
This section consists of a grand total of 24 graphs, so in terms of length it's going to be a long one. But I also think this section is the most interesting one.
So far, we've only looked at maps, their balance stats, and how they lead to aggregate statistics for overall games played. But that leaves a few important questions unanswered. Specifically, one may wonder how much of a role "skill" plays in results, and to what extent these balances can be overcome. And possibly more important than that, balance statistics say nothing about mirror matchups. This last point is particularly important, because the relative skill component of the mirror MUs plays a significant role in one part of my previous argument: that TvT allows the Terran stars to advance more reliably, and allows them to aggregate achievements far better than Z and P can.
As an indicator of overall skill, we will use ELO peak data from the TLPD. It makes sense to look at players from when they were at their strongest to evaluate their skill level, and ELO peak scores are a good way to describe that. We will look at each race individually, and at each matchup individually, to determine how the ELO scores are distributed among the races. In addition, given that there is an uneven number of players by race, we will do the analysis twice: once considering all of the players the race has, and once by considering only the top 50 players of each race by overall ELO peak.
Note that some players in the TLPD did not reach an ELO peak above 2000 - they lost enough games that their final score was below 2000 so a peak was not calculated. For every ELO score where a peak is not calculated, the ELO peak used will be extrapolated as: (current ELO + 2000) / 2.
Let's start by looking at overall ELO peaks by race.
+ Show Spoiler +
For Terran, the average seems to be around 2150, which is pretty solid. For the top 50, it is around 2200, which is very solid indeed. Perhaps most notable from the Terran stats is that there are quite a lot of players who are below average, but also quite a few who are very much above average, with a lot in the 2250+ range. Terran is the only race with anyone above 2400 as well, which is certainly very impressive. All this seems to suggest that Terran lends itself to very strong results, but that many fail to achieve that level and end up below average.
The results for Zerg are actually pretty similar to Terran, with a few interesting differences. Like with Terran, Zerg has an average of around 2150 for all players, 2200 for Top 50. However, the spread of results is more favorable towards the mid-range for Zerg than for Terran. Looking at the Top 50 results, you can see that most players fall well within the 2150-2250 range, with a notable sparsity in the 2250+ range, at least when compared to Terran. This seems broadly consistent with historical results: there are many Zerg players in BW in the "better than average" range, but more Terrans at the very top.
For Protoss, the story is a bit different. The average for all players is around 2150 as well, but the Top 50 range is a bit lower than for Terran or Zerg, at ~2170. It appears that a majority are below the average, but that there is also a sizeable cluster of players in the 2200-2300 range, and a few above 2300 (one above 2350). I don't think it would be too bold to say that these results are noticeably less favorable for Protoss than for Zerg or Terran.
Next, let's look at the vs T matchup for all three races.
+ Show Spoiler +
The TvT average is around 2080, with the Top 50 average being about 2130. Most players are somewhere around average, or quite a bit worse, which suggests that it is a pretty tough matchup to master. Still, there is a sizeable number of players with quite solid scores in the 2200+ range, with one standout score at above 2350.
The ZvT average is ever-so-slightly lower than the TvT one, with a ~2070 average and a ~2130 Top 50 average. There is a pretty big peak slightly below average, but other than that the scores have a distribution that looks a lot like a normal spread in scores. However, it is worth noting that there are far fewer players in the 2200+ range than for TvT, suggesting that that level of mastery is hard to come by. One peak above 2300.
The PvT stats were a little surprising to me. The average is ~2090, and the Top 50 average is ~2120, which is in line with the other data but somewhat underwhelming given the fact that PvT is considered to be a "favored" matchup. The spread of ELO peaks however, is a bit more concerning for Protoss players. While there are a lot of players in the "above average" range, 2100-2150, beyond that there seems to be a lack of Protoss players. Not one breaks 2300, with the best Protoss being around 2250 in PvT. Plus quite a few in the 2000-2050 range, and a few who are well below that. Not really what you would expect for a MU that is supposed to be "favorable" - it seems that there is some inability for PvT players to truly excel at the highest levels of play.
Next, the vs Z matchup for all three races.
+ Show Spoiler +
TvZ looks a little more like what you would expect for a favorable matchup. The average is ~2100, with a Top 50 average of ~2150. The odd thing about these results is simply how scattered they are - there are a lot of peaks, but not at the average. There are a lot of scores at every range, from 1900 to 2350. A lot of players have a very good TvZ peak, with more players in the 2200+ and 2300+ than for ANY vT matchup.
ZvZ is an interesting case. The average is actually pretty high, at ~2090 for all players and a Top 50 of around ~2140. However, perhaps supportive of the "coinflip" assertion of ZvZ play, there is a drastic drop-off in scores past 2200. Most scores are generally clustered around the average, with just a handful that vary significantly. One score is particularly high, in the 2350 range and a good 100 points above the next challenger. That belongs to a certain player who is well known for his particularly impressive JvZ prowess.
PvZ has a ~2080 average and a ~2100 Top 50 average. Not many above 2200, many below average, but mostly a rather normal distribution of results. Again, one player with a very strong vZ score who is famous for being a PvZ star.
Finally, let's look at the vs P matchups.
+ Show Spoiler +
If PvT looks a bit unusual for a favorable matchup, TvP looks mostly standard for an unfavorable one. Average ~2080, Top 50 average ~2120, and a sizeable majority of players are at or below average. However, it's notable that there are quite a few players with very strong TvP, breaking 2200 and even 2300. Remarkably, the TvP ELO is higher than the PvT ELO despite the fact that PvT is supposed to be favored.
ZvP looks like it's mostly normally distributed. Average ~2090, Top 50 average ~2130, with a big cluster of scores near the middle. Not too many players with very high ZvP scores, but there are a handful above 2200 and one above 2300.
PvP is an odd one. Average ~2070, Top 50 average ~2100. A big peak below the average, but there is also a pretty big spread from 1950 to 2200 in which all ELO levels are decently represented. This matchup, like ZvZ, is also sometimes known for its coinflip-like nature, but it's hard to say if that's really true. ZvZ scores are higher on average and have just as many players near the top, but are much more concentrated near middle-range scores than PvP.
Conclusion
So, what does all of this mean? There are a few conclusions that I could come to based on this data:
1. Protoss is decidedly disadvantaged overall.
Looking at everything as a whole - number of players, winrates in favorable and unfavorable matchups, games played, and ELO data - it becomes pretty clear that Protoss players get the short end of the stick. The PvT matchup is not as favorable as ZvP or TvZ, Protoss play fewer games overall, Protoss players have lower ELO peaks, and the Top 50 results for Protoss are noticeably worse than those for Zerg or Terran, meaning that there is a smaller number of top tier Protoss along with fewer Protoss overall. Everything just looks pretty stacked against Protoss.
2. Map balance seems to be consistent with overall results, but it is a small effect.
The map data does not seem particularly shocking when you look at the gameplay results and the ELO results - they all show a slight skew towards the "favored" matchup. But as the "most favorable matchup" graph shows, most maps have at least one matchup that is horribly skewed, which supports the idea that perhaps the variance matters more than the average in map balance. The averages for ELO peaks did not differ all that much by balance - it was the very strong and exceptionally strong scores that differed the most.
3. Zerg has the best middle range of players. Terran has a lot at the top.
This is what the ELO graphs seem to show. All of the Zerg matchups seem to have very solid average and above-average scores, with some that are very strong. The Terran matches differ in their spread, but they all have a lot of players with particularly high scores, in the 2200+ and 2300+ ranges.
4. TvT is the most skill-based mirror, while ZvZ and PvP have a greater coinflip element to them.
A lot of 2200+ scores for TvT suggests that it's a matchup that rewards skill more than PvP or ZvZ, which have very few scores in that range. This factor is perhaps the one that is most blatantly favorable for Terran of all the variables studied.
Overall, it looks like the factors associated with maps, gameplay stats, and ELO generally favor Terran players as a whole. These advantages are small, and often invisible, but they do add up. However, looking only at stats will not be enough to explain how Terran players get such visibly better results, and why all but one of the bonjwas were Terran players. To explain that, we will have to look at the particulars of the tournament structure of individual leagues themselves, which we will do in the sequel.
|
Another great post concluding... what we already knew?
I... can we have a history of these posts? Just lining them up?
And, please don't take this to mean this post was useless.
It is a magnificent post, as I said! Lots of worthwhile effort <3
Now, let us stoke the fires of fiercest argument! I say we nerf T and buff P in Brood War HD *Trollface*
|
Terrans have mostly always been favored with map pools, I think. I felt like in proleague, where coaches can assign maps to a certain player/race, protoss was much more successful. Therefore, I never had any problem with bw's balance. If the most played maps were stuff like judgment day or pathfinder instead of FS or other obvious attempts at giving terrans an advantage during savior's era (reverse LT), I do think the results would be much more different.
|
United Kingdom13775 Posts
On August 28 2016 11:58 Essbee wrote:
Terrans have mostly always been favored with map pools, I think. I felt like in proleague, where coaches can assign maps to a certain player/race, protoss was much more successful. Therefore, I never had any problem with bw's balance. If the most played maps were stuff like judgment day or pathfinder instead of FS or other obvious attempts at giving terrans an advantage during savior's era (reverse LT), I do think the results would be much more different.
It's true; Proleague was where most Protoss players distinguished themselves. But that's basically just saying that Protoss do well when they get to map pick, which is not really the most flattering appraisal of balance considerations.
Do you find Pathfinder to be a good map? Judgment Day was alright, to be sure, but Pathfinder was pretty badly received overall. And for that matter, so was Central Plains, the most blatantly P-favored map in a long time that was so bad that it ended up being all PvP forever.
|
Good read... seems to be what we (roughly) know but it is good to back that up with some statistics...
I guess we can use something other than peak ELO although it is for sure the most convenient stat to use....
|
I love reading stuff like this, so just wanna say thank you op for putting in the effort 
|
Definitely enjoyed this as well. Please keep it up! I'm looking forward to the next installment.
|
Everytime you capitalize ELO, Arpad Elo makes a slight wobble in his grave.
|
Great job! I appreciate the work you have put into this.
|
United Kingdom13775 Posts
On September 01 2016 01:21 opisska wrote:
Everytime you capitalize ELO, Arpad Elo makes a slight wobble in his grave.
ELO = Elo Lowercase Omitted.
On August 31 2016 06:33 HaFnium wrote:
I guess we can use something other than peak ELO although it is for sure the most convenient stat to use....
I think it's about as good as we can get, in terms of both a measure of skill and a usable stat for analysis. It does a good job of measuring relative strength. Its biggest weakness is that there are more games in the modern era than in earlier ones, but that's true for all races so it's not much of a weakness.
|
United Kingdom13775 Posts
A few people have commented about the fact that this writeup covers "basically what we already know" about BW. I'm not sure that's completely true, but I would at least say that it's accurate to say that there's nothing too surprising here. The thing is that this all might not tell you something new, but it is important to establish what's here for what follows.
However, I have to be perfectly honest and acknowledge that the BW scene is not what it once was. I made the original at the nadir of the BW pro scene, and the post-progaming scene has never been anything close to what the progaming scene was even then. It takes a long time to make these, and by the time I'm done there might not be anyone left who is actually interested in Brood War theory-oh. So in the interest of satisfying the curiosity of those who are left who wonder what comes of this analysis, I'm going to summarize it in the form of a statistics-lite post. I already looked at the data and saw what it suggests, but it takes a lot longer to prove that things are as they are than to be able to see it.
The story so far is basically summarized in the conclusion: the maps and the ELO show us that Zerg dominate the mid-range, Terrans the upper range, and Protoss kind of get the short end of the stick. It's also true that Zerg are most numerous, and as a result of their often unreliable mirror matchup they can cannibalize their best players and let their weaker players advance. Terran, they don't necessarily have a lot of players that make it far in the individual leagues but the ones that do, are their very best players. All of the slight advantages that Terran has for top players (highly favorable TvP, TvT, and especially TvZ at the top ranks, abundance of Z players, and highly skilled TvT meaning that the players that are better overall tend to be the ones that advance instead of randoms) add up. What tends to happen is that by sheer quantity Zergs manage to get a lot of titles (most players, just one title each), but Terrans manage to get more titles (and with fewer players) because their best players have all the aforementioned advantages and favorable seeding that allows titles to accumulate into the hands of just a few players - in a way that allows the Terran superstars who are on a tear to be able to cross that "bonjwa" line much easier than Zerg or especially Protoss players can.
In fact, I'd make an even more interesting assertion: the ability to rise to the top hinges most strongly on the ability to beat Zerg players consistently. I think that the high TvZ winrates of all of the Terran superstars is very closely related to their ability to reach the bonjwa line. I also think that Bisu's anomalous PvZ strength is what propelled him so far. I'll also post what I'm going to dub the "Bisu Paradox" from a preliminary blog on this topic.
On May 21 2016 15:37 LegalLord wrote:
There's also another interesting paradox with Bisu, Jaedong, and Flash. Overall, it's pretty clear that Bisu is less accomplished than Jaedong or Flash in the modern era. And yet he's pretty much universally acknowledged to be the best Protoss player in BW history. In terms of achievements, Nal_rA and Jangbi are the only ones who have enough titles, and Stork (and I guess you could make an argument for Reach and Nal_rA, but it's a longshot) is the only one who is close enough on skill level, to even be in the position to contest that dominance. Jangbi was decidedly inconsistent throughout his years and just managed to put together two titles at the very end, so I'd say he's out. Stork is hard to argue for since he only has a single title, so he's also out. Nal_rA is probably the closest, but Bisu had a more dominant run and has more titles to his name, and I think it would be a hard sell to say that Bisu isn't the better player. For Jaedong, he has Savior to contend with and that's a tough one to evaluate (personally I can't really choose between the two). Flash, he has NaDa and iloveoov as pretty solid challengers for the title of best Terran. Curious, to be sure.
Finally, there's a really cool graphic on the wiki page for the OSL that describes a lot of relevant statistics: http://wiki.teamliquid.net/starcraft/OnGameNet_Starleague_(OSL)
Now that's a whole lot of assertions in a pretty short post. I'm pretty sure that they're accurate, and I'd say that that is a lot more substantial and surprising a claim than what you see here - but also it needs what I posted here to even look valid. It'll just take a while to actually run all of the numbers into a full analysis and to write a solid post on the topic, and the next few months look really busy for me so I can't promise that I will be able to manage before the BW scene degrades to the point that there won't be anyone left to write for.
I was considering a fourth post, which would be a simulation designed to show what variations in skill and tournament structures tend to do to results, but I'm definitely cancelling that one. I just don't have the time and I don't think it would really do all that much more to prove my point.
So for the few of you who are wondering where this all goes, there you go.
|
There has been similar posts on other sites as well.
http://www.pgr21.net/pb/pb.php?id=free2&no=41894&category=1&divpage=11&sn=on&keyword=信主SUNNY
This was one of the higher quality posts on PGR21, a Korean site. It took the results of dual-major individual league era (post 2002) post-round of eight, because the seeding and results of the round of sixteen and below tend to be heavily influenced by the previous league, and the post-round of eight results are almost purely indicative of the performances of the players participating in that specific league. In case you cannot read Korean, blue represents the terran race, red represents the zerg race, and green represents the protoss race.
The terran race as a whole rarely has a particularly bad time in the post-round of eight stages, except when the best respective periods for the protoss, then the zerg race circa 2008~2009, happened back to back. However, their representation tends to be more steady than the other two races, who tends to be more feast or famine with their representation.
You can also see which players carried the torch for each of the races during their darkest times, which is YellOw circa 02~03, Reach circa 02~05, and FanTaSy circa 08~09.
Just like you asserted, historically speaking, terran players tend to dominate the latter stages of the individual leagues, while protoss players got the short end of the stick.
There is also a very good post on the map balance of various maps in history.
http://pgr21.com/pb/pb.php?id=free2&no=55995
Out of the 229 maps he analyzed, there were 32 maps that had a higher than 60% predicted win rate for the terran race, and the same number for the zerg race. However, the difference is that while the bad maps (Geometry, Othello, and Neo Moon Glaive in particular) that favour the terran race tends to be more recent, the bad maps (with the exception of Battle Royal) that favour the zerg race tends to be from the very ancient pre-1.08 patch era (such as Legacy of Char, and Ashrigo). The protoss race had only eight maps that had a higher than 60% predicted win rate.
As you mentioned, four player maps are the most played map, and the terran race benefits the most from this pattern, as they can maximize their domination over the zerg race, and actually reverse their usual disadvantage towards the protoss race to their favour.
Terran players struggle the most on three player maps as they struggle harder against both the zerg and the protoss race compared to other two or four player maps. However, this isn't a huge issue as three player maps are the least played maps.
The problem is that, once you use three player maps to lessen the statistical advantage terran players have over the other two races, you have to further damage what is already statistically the most imbalanced match-up, the zerg versus protoss match-up.
Two player maps seems to be easier to balance than the other two kind of maps, but I guess the relative repetitive nature of these maps must have hurt the viewership, as well as the lack of "fighting" for the remaining starting positions tends to make the games rather passive in game play.
I mean, discussion on balance, and maps are a never ending story, but while the specifics are debatable, and exact approach differs from people to people, the overall consensus remains the same. The terran race enjoyed the most success in the professional scene partly due to the 1.08 patch buff, as well as the rise of Lost Temple-style four player maps, while the protoss race got shafted compared to the other two races.
|
Outside of the statistics, as a person who watched pro broodwar from the very beginning (I've probably seen 90% + of all broadcasted games until the last OSL). I consider Terran's imbalance to come mostly from the higher tolerance that the race has towards the mid layer human mistakes/plays.
To be more clear if we divide the plays a player can make in 3 layers - low, mid and high. Where:
1). Low play is when a player risks little and if successful earns little, but if unsuccessful loses little.
2). Mid play is when you risk average to gain or lose average advantage
3). High play is when you somehow all in for either win or lose
In case 1) usually all three races can turtle in and stay in the game, regardless who made the play and the outcome.
In case 3) usually the game ends for each race, if it works you win, if not you lose, regardless of the race
However, according to my observations case 2) is where Terran shines. This where Terran's successful plays often leads to case 3) where they win the game, but, the unsuccessful plays can often be mitigated to case 1) where Terran can stay into the game.
For example a successful one dropship M&M drop can hurt Zerg in a way that it will lose in next few minutes, while losing the dropship sets the terran only slightly behind.
Vultures run by can lead to an easy win vs Protoss, while losing the vultures don't hurt as much.
And this goes the opposite direction as well. If a zerg 3rd get delayed or protoss reaver/dark surprising attack do not do enough damage they often cannot sustain into the game long after.
Thus, in long steady games, the races look pretty balanced, we see everything being used and we see that each race has its tools to win, we see how some players are just better than others regardless of the race, but Terran can sneak more of these "easy" wins and skew the statistics. Which may mean that Terran is actually OP, but maybe not. Depends how you look at it.
|
United Kingdom13775 Posts
On September 01 2016 16:40 Letmelose wrote:There has been similar posts on other sites as well. http://www.pgr21.net/pb/pb.php?id=free2&no=41894&category=1&divpage=11&sn=on&keyword=信主SUNNYThis was one of the higher quality posts on PGR21, a Korean site. It took the results of dual-major individual league era (post 2002) post-round of eight, because the seeding and results of the round of sixteen and below tend to be heavily influenced by the previous league, and the post-round of eight results are almost purely indicative of the performances of the players participating in that specific league. In case you cannot read Korean, blue represents the terran race, red represents the zerg race, and green represents the protoss race. The terran race as a whole rarely has a particularly bad time in the post-round of eight stages, except when the best respective periods for the protoss, then the zerg race circa 2008~2009, happened back to back. However, their representation tends to be more steady than the other two races, who tends to be more feast or famine with their representation. You can also see which players carried the torch for each of the races during their darkest times, which is YellOw circa 02~03, Reach circa 02~05, and FanTaSy circa 08~09. Just like you asserted, historically speaking, terran players tend to dominate the latter stages of the individual leagues, while protoss players got the short end of the stick. There is also a very good post on the map balance of various maps in history. http://pgr21.com/pb/pb.php?id=free2&no=55995Out of the 229 maps he analyzed, there were 32 maps that had a higher than 60% predicted win rate for the terran race, and the same number for the zerg race. However, the difference is that while the bad maps (Geometry, Othello, and Neo Moon Glaive in particular) that favour the terran race tends to be more recent, the bad maps (with the exception of Battle Royal) that favour the zerg race tends to be from the very ancient pre-1.08 patch era (such as Legacy of Char, and Ashrigo). The protoss race had only eight maps that had a higher than 60% predicted win rate. As you mentioned, four player maps are the most played map, and the terran race benefits the most from this pattern, as they can maximize their domination over the zerg race, and actually reverse their usual disadvantage towards the protoss race to their favour. Terran players struggle the most on three player maps as they struggle harder against both the zerg and the protoss race compared to other two or four player maps. However, this isn't a huge issue as three player maps are the least played maps. The problem is that, once you use three player maps to lessen the statistical advantage terran players have over the other two races, you have to further damage what is already statistically the most imbalanced match-up, the zerg versus protoss match-up. Two player maps seems to be easier to balance than the other two kind of maps, but I guess the relative repetitive nature of these maps must have hurt the viewership, as well as the lack of "fighting" for the remaining starting positions tends to make the games rather passive in game play. I mean, discussion on balance, and maps are a never ending story, but while the specifics are debatable, and exact approach differs from people to people, the overall consensus remains the same. The terran race enjoyed the most success in the professional scene partly due to the 1.08 patch buff, as well as the rise of Lost Temple-style four player maps, while the protoss race got shafted compared to the other two races.
Interesting that the Korean world has looked at something similar - and in 2015 no less. And as you said, the approach may be different but the results are mostly the same, lol.
My personal opinion is that it's best to look just at league winners - not at finals, semis, or Ro8. It's a little narrow a criteria, but it's the only one that is hard to dispute as valid. If you have Flash and 31 random D level players from ICCup in a tournament together, 7 of those D levels will make the top 8. One will get a silver. And yet the winner is on a whole other level than the next players. A more real example is WCG International, where #4 is merely the best player who managed to avoid the three Korean players for the longest - even if said player is far from the 4th best player in the tourney. Because while there are players who were better and who were worse who managed to win them (and ones that got uniquely favorable series on their way up), they still managed to beat every single opponent in the running for their title. I can't say the same for silvers - I saw plenty of weak players get silver medals because they fumbled their way into the big leagues.
Maps... yeah they somehow end up becoming Terran favored overall. The reasons why are pretty hard to analyze but the results speak for themselves. I just think that a lot of explicitly Protoss maps tend to get culled pretty quickly or involve unfair mechanics like island maps. Whatever the reason, the outcome is clear to see: post-1.08 maps have a Terran-favorable flavor pretty consistently.
|
United Kingdom13775 Posts
On September 02 2016 03:44 M2 wrote:
Outside of the statistics, as a person who watched pro broodwar from the very beginning (I've probably seen 90% + of all broadcasted games until the last OSL). I consider Terran's imbalance to come mostly from the higher tolerance that the race has towards the mid layer human mistakes/plays.
To be more clear if we divide the plays a player can make in 3 layers - low, mid and high. Where:
1). Low play is when a player risks little and if successful earns little, but if unsuccessful loses little.
2). Mid play is when you risk average to gain or lose average advantage
3). High play is when you somehow all in for either win or lose
In case 1) usually all three races can turtle in and stay in the game, regardless who made the play and the outcome.
In case 3) usually the game ends for each race, if it works you win, if not you lose, regardless of the race
However, according to my observations case 2) is where Terran shines. This where Terran's successful plays often leads to case 3) where they win the game, but, the unsuccessful plays can often be mitigated to case 1) where Terran can stay into the game.
For example a successful one dropship M&M drop can hurt Zerg in a way that it will lose in next few minutes, while losing the dropship sets the terran only slightly behind.
Vultures run by can lead to an easy win vs Protoss, while losing the vultures don't hurt as much.
And this goes the opposite direction as well. If a zerg 3rd get delayed or protoss reaver/dark surprising attack do not do enough damage they often cannot sustain into the game long after.
Thus, in long steady games, the races look pretty balanced, we see everything being used and we see that each race has its tools to win, we see how some players are just better than others regardless of the race, but Terran can sneak more of these "easy" wins and skew the statistics. Which may mean that Terran is actually OP, but maybe not. Depends how you look at it.
A bit of a rare sighting, cool.
I do generally agree with your conclusion about how Terran gets its advantage. I'll admit that I was not a BW player who was of the caliber to be able to evaluate this - and it's been so long that I've lost a lot of the touch I once had - but I definitely saw a much more promising consistency within Terran than with other races when I was playing. A lot more decisions that were risky led to a "recovery mode" position rather than an "instant loss" position with Z or P. It might even be reasonable to make the rather aggressive assertion that what qualifies as a (3) for Zerg or Protoss get to be a (2) for Terran - risky moves that are potent and can end the game, but that have less commitment than Z or P attacks that become a de facto all-in because of how they affect the economy. For a lot of the titles that Flash won, he abused the hell out of (2) style moves that really just exploited the Zerg opponents he beat (Jaedong and Zero) in usual ways that gave him a powerful recovery/followup while being aggressive enough to be able to end the game.
So, yeah. Terrans get an advantage in the form of strong consistency. Z and P can win or lose but T is positioned to be able to abuse their recovery abilities to make (2) moves that can have a (3)-like character to them. Or at least, so I have seen for the few years before 2012 that I've been following pro BW.
|
United Kingdom13775 Posts
Bump after two months with a slight update.
So I mentioned that I wasn't sure if I was going to make a sequel. At this point I think it'd be more work than I really have time for. But the story as I told it is far from complete, and I do want to add more to it. So what I'm going to do is just post occasional addenda to this thread, for those few people who care enough to read through some very long BW theory-oh. That gives more content to look at, but also takes much less time in the form of having to sit down and write the entire piece.
So today's addendum is a brief analysis of MSL versus OSL to analyze the somewhat grandiose statement made at some point in the past that "OSL is where champions are made, MSL is where random people have a chance to prosper." That's a bold and likely hard to verify claim. I was originally planning to make a simulation to test the validity of this claim, but I'm definitely not doing that. However, by inspection of historical results we can get a pretty good picture of what the reality is. Keep in mind, there have been 26 completed MSLs and 34 completed OSLs, roughly a third more OSLs. The full results are detailed in the Liquipedia links above.
MSL
Number of players with
3 Golds: 5
2 Golds: 1
1 Gold: 9
OSL
Number of players with
3 Golds: 4
2 Golds: 4
1 Gold: 14
I have generally made the (somewhat contentious) claim that focusing on gold medals and ignoring silver medalists in the analysis is the best choice. The simple reason is that there are far more "fluke silvers" than "fluke golds" and I think that that is an important thing to consider. I'd be open to a counterargument however. Nevertheless, I'm focusing on golds only.
So there are more triple-golds in the MSL than in the OSL, even without scaling. There are six double plus triple golds for the MSL, and eight for OSLs, which is roughly the same to scale. Similarly, the 14 single winners is a little more than the 9 (12 after scaling) of MSL. So the MSL is biased towards fewer unique winners and more medalists in the hands of a few players.
Last thing we're going to look at is back-to-back victories. We're going to look at double-wins (back-to-back win) and triple-wins (back-to-back-to-back win). A triple counts as two double wins. In addition, we're going to slightly relax the conditions for both, and allow almost back-to-backs to count as well. For example, if someone won an OSL, lost the next one, then won the one after that, that would be a relaxed double win. If they further lost one more after that then won a third one, that is a triple win and two double wins. The results:
MSL Strict Conditions:
Double wins: 7
Triple wins: 2
MSL Relaxed Conditions:
Double wins: 9
Triple wins: 4 (all the triple winners except Bisu, lol)
OSL Strict Conditions:
Double wins: 3
Triple wins: 0
OSL Relaxed Conditions:
Double wins: 4 (only added one is Flash, where he was blocked by Effort in between two wins)
Triple wins: 0
Well, that result is pretty telling. We can see that the MSL clearly lends itself better to long streaks and to winning many titles. There are far more random victories in the OSL than in the MSL. Looks like the hype statement in regards to the OSL doesn't really hold much water.
|
|
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/109Hwts.png)
![[image loading]](http://i.imgur.com/YGjtRMT.png)
![[image loading]](http://i.imgur.com/DV3yzgb.png)
![[image loading]](http://i.imgur.com/5g74why.png)
![[image loading]](http://i.imgur.com/7iPLLbw.png)
![[image loading]](http://i.imgur.com/s95OtGB.png)
![[image loading]](http://i.imgur.com/AMhv6Mp.png)
![[image loading]](http://i.imgur.com/ZEYim9s.png)
![[image loading]](http://i.imgur.com/GvB4Q4w.png)
![[image loading]](http://i.imgur.com/nkHl6ZT.png)
![[image loading]](http://i.imgur.com/QWJJVK0.png)
![[image loading]](http://i.imgur.com/JutpYDm.png)
![[image loading]](http://i.imgur.com/fXx0Etj.png)
![[image loading]](http://i.imgur.com/pDFnw5x.png)
![[image loading]](http://i.imgur.com/Z5mgw91.png)
![[image loading]](http://i.imgur.com/Ur8Vegc.png)
![[image loading]](http://i.imgur.com/In2OfLP.png)
![[image loading]](http://i.imgur.com/PLRjyTa.png)
![[image loading]](http://i.imgur.com/r1VfZIf.png)
![[image loading]](http://i.imgur.com/594eUdN.png)
![[image loading]](http://i.imgur.com/lzXDgwj.png)
![[image loading]](http://i.imgur.com/SFLYOiR.png)
![[image loading]](http://i.imgur.com/qnOz04c.png)
![[image loading]](http://i.imgur.com/R7HPMtn.png)
![[image loading]](http://i.imgur.com/Q5vXOhk.png)
![[image loading]](http://i.imgur.com/nPBqA8B.png)
![[image loading]](http://i.imgur.com/aAmsIjt.png)
![[image loading]](http://i.imgur.com/VqIR6V2.png)
![[image loading]](http://i.imgur.com/zw7FRGB.png)
![[image loading]](http://i.imgur.com/hoBATju.png)
![[image loading]](http://i.imgur.com/aI3NFCX.png)
![[image loading]](http://i.imgur.com/Y0AbMss.png)
![[image loading]](http://i.imgur.com/8a3KlRh.png)
![[image loading]](http://i.imgur.com/1JnqDnq.png)
![[image loading]](http://i.imgur.com/N3vLSwB.png)
![[image loading]](http://i.imgur.com/39PyRbk.png)
![[image loading]](http://i.imgur.com/2dWY57x.png)
![[image loading]](http://i.imgur.com/E2WltXY.png)
