I appreciate you getting this data but I'll be real:Ok so basically this is kind of late but
I wanted to take a more objective and macro-level look at the impacts of Terastalization, so I've spent the past week and a half or so writing up some code that can take in a replay and output a set of stats indicating the performance of each Pokemon in that game, such as the number of active turns, times brought in, KOs, % damage dealt, and more. More importantly, I also tracked which Pokemon terastallized in a game, enabling us to compare the performance of pokemon that terastallize to those that do not to see if we get any significant results.
Turns Active specifically refers to how many times a Pokemon used a move during battle.
Times Brought is pretty self-explanatory; this includes being the lead, being pivoted in, coming in after a sack, and just raw switching in.
Percent Damage Dealt only includes damage dealt by direct attacks. Chip dealt by hazards, poison, salt cure, etc. are not included in this variable. If a pokemon damages a Substitute that damage is not added to this variable.
KOs is the number of opposing pokemon that were knocked out while the pokemon was active, including damage from direct attacks and indirect damage as well. If an opposing pokemon knocks itself out with Healing Wish, Explosion, etc. this does not count as a KO.
Hits Taken is the number of direct attacks a pokemon sustained, as well as the number of times the pokemon is swapped into a move it is immune to. This aims to be the main metric by which a pokemon's defensive capabilities are measured.
Percent Healed sums all healing done to a pokemon, including passive recovery from Leftovers. If another pokemon passes Wish/Healing wish to the pokemon, the pokemon that received the wish gets its percent healed stat added to.
The Tera indicator variable is a 1 if the pokemon terastallizes during a game, and is 0 otherwise. If a pokemon does terastallize, its stats reflect its overall performance in the game, both before and after terastallizing.
For my dataset I used every SV WCoP game after the Urshifu and Volcorona bans, giving a dataset of 232 games to work with; the dataset can be found here. I also dropped all pokemon that had 0 active turns in battle, thus wouldn't even have the chance to tera if it wanted to. For my analysis I did a fixed-effect regression using terastallization as the explanatory variable against several different dependent variables. Simply put, for each pokemon it takes the difference between its performance without terastallization and compares it to its performance with terastallization to compute the overall effect terastallization has on pokemon performance. I also looked at how terastallization impacts the performance of the most commonly terastallized pokemon in WCoP, being Kingambit, Garganacl, Iron Valiant, and Baxcalibur. In light of the incoming Kingambit suspect I also ran the regression on all pokemon excluding Kingambit, to estimate the impacts of Tera in a metagame without Kingambit. The results can be found here.
Surprisingly, terastallization had significant results for most metrics I chose to look at, meaning we can be confident that these results weren't just caused by random chance in our dataset. Notably, on average, terastallization increases the KOs a pokemon attains by 0.679, the % damage dealt by 61.984, hits taken by 0.468, and the number of active turns by 1.642, indicating that terastallization has nontrivial impacts on a pokemon's ability to make progress in a game. However, a pokemon that terastalizes does not have significant effects on the number of times it switches in, supporting the premise that terastallization is mainly utilized for short-term swings and endgame cleanups than to improve a pokemon's longevity over the course of a game.
Looking at some pokemon individually, we first see that Kingambit yields results mostly similar to that of Tera on the entire dataset. From this lens it seems that terastallization may not be as major of a factor in Kingambit's current power levels.
Garganacl's results are also not particularly surprising; Tera lets it sustain more hits and Recover more often than the average Tera user. Offensively, its numbers are rather low since Salt Cure isn't being added to damage numbers currently.
The most surprising results come from Iron Valiant and Baxcalibur's performances. Both of these pokemon don't have significant results for hits taken, but have the highest significant numbers for damage dealt and KOs, showing their ability to suddenly Tera to net a KO during a game rather than utilizing Tera defensively for setup opportunities.
Lastly, the results of the analysis without Kingambit in the dataset are mostly similar to that with Kingambit included, so part of me is skeptical how much a Kingambit ban will actually do for making terastallization healthier.
One thing to note however, these regressions also resulted in low R^2 values across the board, meaning that the effects that terastallization has high variance on the outcome. While this high variance may at first glance imply the inconsistency of the mechanic and the ability to play around it, a look at the actual distributions of our variables argues an opposite story. For instance, take a look at the difference between the number of KOs attained by pokemon that do not Terastallize versus pokemon that do:
View attachment 537631View attachment 537632
Here, we see that despite high variance, the ability of terastallization to net a pokemon KOs is quite consistent, with less than 30% of terastallized pokemon obtaining no KOs at all. This compares to over 50% of non-Tera pokemon getting no KOs in a game. The high variance instead comes into play through a terastallized pokemon's ability to net potentially several KOs in a game. The consistency at which this occurs- more than 70% of terastallized pokemon attain a KO in a game- demonstrates that the mechanic currently is very difficult to play around, even at a tournament level. Because of this, I'm leaning to support some sort of tiering action for Tera. So yeah let me know if you have any feedback or questions or other things I should analyze
This data shows literally nothing about Tera being bad or good for the game, because it will usually be the wincon anyways;
I heavily encourage you to keep doing things like this though, especially for suspects.
Last edited: