A Bradford-Hill analysis of the APM/win-rate correlation:
- Strength. There must be very low odds of occurring by chance. APMs correlation with mmr is 0.65 and there is a statistically significant correspondence between lower APM for the same performance level within the Protoss group (237 p vs 290 t/z in gm).
- Consistency. Findings of the above correlation must be consistent across multiple sources. Data from sc2replaystats, blizzard’s own employees, and a variety of other sources are in agreement that the correlation between winrate and APM is very large, but lower within the protoss grouping.
- Specificity. A factor is likely to be causative of an outcome if the outcome occurs within a specific population with no reasonable explanation other than involvement of the factor in question. Protoss in GM score lower in every known skill metric ranging from apm to supply blocks to spending quotient to screen movements; the only possible explanation is that it requires fewer substantive tasks within the game, as protoss, to achieve the same rank as a terran or zerg. This relationship is especially stark when contrasting the effects on pro level play, where high level protoss often forget key upgrades or lose multiple oracles due to unforced errors that are APM-related.
- Temporality. The outcome must occur chronologically after the cause. The actions inputted into the game cause the win or the loss of that game which affects ranking and so the substantive actions within the game clearly have a temporal relationship with the probability to win.
- Gradient. The relationship between the cause and the effect should be proportional, meaning if the cause increases a little then the effect increases a little as well; if the cause increases a lot then the effect increases a lot. APM and winrate have a very proportional relationship all the way from bronze up to serral.
- Plausibility. The association must have a known mechanism that causes the effect. Obviously being able to do more tasks within a game where time is the most important resource results in better performance, thus higher APM causing higher performance is plausible.
- Coherence. The interpretation of the relationship cannot conflict with what is already known about the relationships. According to industry experts ranging from the creators of Command and Conquer to the lead designer of Supreme Commander, “The manual dexterity and ability to multitask and divide one’s attention is often considered the most important aspect to succeeding at the RTS genre”. Additionally, the AlphaStar research team found that Protoss achieved the highest performance with the lowest APM, thus recreating the relationships observed in ladder data, which is, in effect, a defacto experimental study. Clearly this theory is in coherence with industry experts.
Translation, the 0.65 correlation between APM and win-rate meets the Bradford-hill criteria for causality, and thus APM causes the majority of SC2 performance. For a given skill level, such as Grandmaster, Protoss are lower in skill metrics (like APM) and this is proof that lower skill achieves the same performance level, which is analogous to saying Protoss is overpowered. This effect is so extreme that Protoss in GM have been measured to have lower APM metrics than Masters-level Zergs, which suggests the effect that Protoss has on performance is rather extreme.
Possible explanations:
- High supply cost units. Protoss armies are made of fewer units which are bulkier and more durable. This means fewer units to build, fewer units to micro, fewer units to position and maneuver. The durability increases the time-window where an action can take place to save the unit with micro. This obviously reduces the multitasking requirements of the race, making it less likely that they make multitasking-related errors.
- Units that are designed to have no micro, such as zealots and their “auto-charge” ability. Remove the auto-cast from charge, make charge have to be casted on a target enemy unit, so that it operates similar to blink, and we’d probably see an increase in the difficulty required to play protoss.
- Plain old overpowered. Carriers and storms have a very high power level, and this makes the units easy to use. It’s very hard to mess up when messing up still results in a favorable outcome due to the sheer strength of the abilities involved.
How do we fix this:
- Stop listening to protoss whiners who cry about “muh serral wins muh premier tournament”. Clearly the pro level performance is a skill issue. Their whining is not going to help solve the real issue (which is skill).
- Prioritize the ladder gaming experience over the crybabies in the pro scene. The actual players of the actual game are far more important to the survival of the game than the elitists at the top. The elitists at the top are professionals who are paid to solve the game problems with their absurd skill; the idea that we cater balance to them is absurd when the entire premise of their profession is the polar opposite of what is healthy for the game.
- Increase the micro and multitasking of Protoss units until the correlation between APM (and other skill metrics) and MMR is roughly equal to what it is for Terran and Zerg. Remove auto-cast from abilities like charge. Reduce the power level of carriers and storms. Remove auto-micro from the interceptors / make protoss have to use micro on the individual interceptors. Etc.
- Decrease the multitasking requirements of terran and zerg. Add auto-cast to the inject ability, for example.
THANK YOU FOR ATTENDING MY TED TALK
(and please place your counter-arguments in the trash bin on your way out)
BONUS (LMFAO): https://i.imgur.com/7iF8S7R.jpeg