Season Eight Fantasy Retrospective

Unlike the blowout victory for Amicalola in Season Seven, this year instead featured an insanely tight race where six different fantasy teams had the chance to take the crown going into the final match. Gandhi's second place finish narrowly boosted Kjotleik to the victory over Bellomorphe's late-charging Elizabeth with j_mie6 a single point back and antisocialmunky's team probably being the one that "should have" won given how dominant Mansa Musa proved to be in the alternate histories for the championship game. Eauxps and Amicalola were the only competitors who were truly out of the running and even they looked like they could have had a real shot if a few of the matches had played out differently. This was by far the most exciting fantasy season we've had thus far and it was a ton of fun following along throughout the season.

After the season concluded and the community started running the alternate histories, I wanted to go back and revist the fantasy contest to take a fresh look at which draft picks turned out to be the best choices in light of running those replays of each map. For example, we know that Gandhi was exceptionally lucky to make it to the championship in one piece while Cyrus somehow managed to avoid scoring any points in Game Eight despite nearly reaching 100 points in the alternate histories. How would things have played out in the fantasy contest if we'd run these matches multiple times? We're therefore going to take a look at each of the opening round games in the specific context of the fantasy draft using the additional data from the alternate histories. Note that this analysis can *ONLY* be done on the opening round games because naturally the playoff matches were dependent on what happened in the first eight games. If a different leader had been victorious in Game One then there would have been different leaders in Playoff Game One, and so on. Similarly, this analysis will naturally undervalue heavy bidding on big favorites because it only counts the scoring from the opening round games. Dropping a heavy investment into Huayna Capac or Justinian is all about getting value from more than a single game and it can absolutely be worthwhile to overbid with the expectation of paying for future playoff performances. So with those caveats in mind, let's take a look at the opening round matches starting with Game One:

Game One from this season will go down in AI Survivor history for having one of the most improbable results ever seen on Livestream, as Hammurabi won outright despite having a 90% fatality rate and never coming even remotely close to a win in the alternate histories. This bizarre outcome meant that several excellent leader bids in the fantasy contest ended up getting wasted, especially the 17 gold bid from Henrik on Peter which in retrospect looks like an absolute steal. Hammurabi scored 8 actual points for El Grillo and Churchill similarly earned 9 fantasy points in the real competition, however these were low-odds results that were almost never repeated in the alternate histories. J_mie's bid of 35 gold for Churchill was in truth a pretty serious overspend that happened to work out by dumb luck. And unfortunately Boudica was apparently a bad choice at any price, not even having the courtesy to be worth many First to Die points for antisocialmunky. This was a really wacky outcome that did not in any way match the normal results of Game One and it threw off the fantasy contest right from the start.

In stark contrast to Game One, Game Two played out very much as expected in the Livestream match. The alternate histories confirmed that Justinian and Gilgamesh were the top two leaders on this map which was exactly how they finished in the alternate histories, aside from Gilgamesh doing notably better in most of the map replays. Both of these leaders were expensive bids for Kjotleik's fantasy team and they both delivered in their opening round match (before going on to falter in the playoffs). Most of the other leaders didn't attract much interest in the fantasy auction and were taken as longshot odds at cheap prices, with Wang Kon and Bismarck holding the best value due to their First to Die chances. Tokugawa dying first in the actual Game Two was pretty rare but he was a cheap enough purchase at 17 gold that it was worthwhile for Euaxps to take a flier on his chances. The one truly unfortunate purchase from this game was El Grillo spending 31 gold on Genghis Khan - due to an aggressive bidding war agains Amicalola! - who turned out to be much weaker than everyone expected. The Mongol leader never sniffed a top two finish in any of the alternate histories and wasn't even a good First to Die candidate either.

After the brief return to sanity in Game Two, we were once again back to more low-odds craziness in the Livestream version of Game Three. This was the shocking match where Sitting Bull took first place and Frederick took second, with the only two high peaceweight AI leaders somehow advancing against their four low peaceweight competitors. Needless to say, the alternate histories confirmed that this was an outlier result with Willem and Huayna Capac being the most likely leaders to prevail. The weird outcome resulted in Euxaps and j_mie getting tremendous value out of their respective Frederick and Sitting Bull purchases, bids that looked a lot worse when taking the alternate histories into account. Sitting Bull was definitely worth 18 gold since he was First to Die so often but 24 gold for Frederick, the lowest-scoring leader in the alternate histories, probably was a misstep that got lucky. Conversely, Amicalola had a tremendous value bid on Willem that failed to work out and Bellomorphe's expensive purchase of Huayna Capac was nowhere near as bad as it looked at first. The worst bid here was 64 gold for Shaka by El Grillo as the AI Survivor community collectively misread his starting position. Grillo was hardly alone in thinking that Shaka was likely to take a top two finish from this position but it failed to play out that way in the real Game Three or in the alternate histories.

Game Four was a very straightforward game to explain both in terms of the actual performance and in terms of the fantasy contest. Mansa Musa was seen as a prohibitive favorite, antisocialmunky submitted the highest bid in the whole fantasy auction to get him on his team at 86 gold, and then Mansa backed up that confidence by winning the real Game Four along with ALL TWENTY map replays. I think it wasn't until the 27th replay of the map that there was finally a result where Mansa didn't win, sheesh! Due to the high cost of purchasing Mansa and his low-scoring victories (1 kill in 20 games?!), there's a case to be made here that Mansa wasn't that great of a purchase, especially not knowing ahead of time that this would wind up being the season of the high peaceweight leaders where Mansa would have a diplomatic advantage in every playoff game. In any case, this purchase certainly worked out for antisocialmunky and very nearly delivered him the fantasy crown had it not been for an unlucky raze of a Malinese city by a wandering barbarian in the championship game.

Among the other non-Mansa leaders in this game, there was a medium-sized bid on Saladin from Euaxps that failed to work in the real Game Four (with Saladin falling victim to First to Die status) but looked better in the context of the alternate histories. Saladin was clearly the best of the non-Mansa leaders here even if that wasn't saying too much. The best bid was probably antisocialmunky getting Pericles for 19 gold who performed much better in the alternate histories than in the pathetic performace we saw on Livestream. The worst bid was definitely 30 gold for Qin Shi Huang from Bellomorphe, which looked great when Qin finished second on Livestream but which turned out to be an unlikely outlier result. Qin's 9 fantasy points across the alternate histories was the single lowest-scoring result of any leader in the opening round which was certainly not worth the cost he required. Honestly though, Mansa was so dominant in this game that none of the other leaders were particularly good bids, and Mansa's fast Cultural victories meant few kills for anyone else and thus fewer chances to score fantasy points. This was a "stay away" game for fantasy scoring purposes that benefited everyone not spending on these leaders.

Game Five was another match where the Livestream results and the alternate histories diverged sharply from one another. We watched a game where Ramesses was able to sneak out a victory from the western seaboard, only to see Ramesses fail to land even a single win across the 20 map replays. Instead it was Pacal who was the truly dominant leader for this setup and thus Kjotleik's bid of 25 gold was a truly excellent purchase. Henrik was also able to pick up Hatshepsut at a cheap 17 gold pricetag only to see the Egyptian leader once again fail to score a single point in dramatic fashion. Hatty has been almost comically unlucky across several of these AI Survivor seasons and Henrik's fantasy team was just as snakebitten as many smart auction purchases flopped in the Livestream games. Brennus' high kill count made him a solid value for El Grillo's team in the alternate histories and then Brennus later had an awesome Wildcard game to score more points in the real fantasy competition. The worst purchase by far was Darius going for 45 gold to antisocialmunky as the Persian leader was a true flop on this map. I was very much fooled myself since I thought he was going to be the favorite on this map and even had to run emergency alternate histories that proved he was instead a big sucker - whoops!

This was a complete mess of a game where even the alternate histories couldn't make much sense of what was happening on the map. Game Six was a high variance game with a whole bunch of different possible outcomes as every single one of the seven leaders took at least one victory in the alternate histories. Stalin was the best leader with the most victories and the most kills (though still only winning about a third of the time) as Henrik was once again robbed of an excellent fantasy purchase in the Livestream game. Getting Stalin for only 30 gold was great value that unfortunately didn't work out. Then Henrik also had a good purchase of Joao, the second-strongest performing leader in Game Six, which yielded only a single fantasy point from a kill scored later by the Portuguese leader. Amicalola was in the same boat with a really good bid of 15 gold on Roosevelt that only turned into 2 fantasy points (both from kills!) as Roosevelt picked a bad lategame fight with Victoria. The actual points scored from Game Six went to Victoria and Gandhi... who were the two lowest-scoring leaders on this map in the alternate histories! Kudos to Euxaps and Kjotleik for making cheap value purchases that managed to hit in the Livestream game but this was very much a case of RNG at its finest. It's hard not to feel for Henrik and Amicalola who were both robbed in the real Game Six.

And speaking of another game where someone was robbed, Henrik once again submitted a cheap bid of 21 gold on the best-performing leader only to see Mehmed finish in a distant second place to Elizabeth in the actual Game Seven. Mehmed had the most kills and won nearly as many games as every other leader combined but couldn't quite make it happen on the Livestream. Antisocialmunky was also quite unfortunate to see Charlemagne collapse in the Livestream game and only score First to Die points while he was one of the better-performing leaders on this map in the alternate histories. Bellomorphe was a bit unlucky not to score any fantasy points for Isabella, however this was more than counterbalanced when his longshot Elizabeth bid took an unlikely first place en route to 13 total fantasy points and the Season Seven Championship. The one genuinely bad purchase from this game was Suryavarman going for 50 gold to Kjotleik as practically everyone in the community thought that the Khmer would be really strong on this map. I was completely fooled by this as well and never expected Suryavarman to score nearly as many alternate histories points from First to Die status as from top two finishes. Finally, there was some speculation that Montezuma would do something on this map but, being Monty, that never materialized in ther alternate histories either.

Our last opening round game had yet another curveball result as Cyrus proved to be extremely dominant in the alternate histories but without being able to claim either top two spot on Livestream. J_mie made an expensive bid of 56 gold for Cyrus that should have paid off with a big fantasy score and which simply didn't happen due to bad luck. Most of the other leaders on this map were tightly bunched together in terms of alternate histories scoring, clustered right around 30 fantasy points. This meant that Suleiman for 12 gold and Lincoln for 13 gold were both superb value purchases for Euxaps, and he managed to hit on one of them when Lincoln somehow ended up with a second place finish on Livestream. 33 gold for Louis was defensible by Amicalola but didn't wind up hitting in practice. The only really bad purchase was 50 gold for Julius Caesar, also on Euaxps' fantasy team, as the western Roman leader utterly failed to get going in the real Game Eight or in the alternate histories. The last few seasons truly have not been kind to Caesar ever since we removed the extra Deity starting techs.

Here's the combined fantasy summary from the alternate histories of all eight opening round games:

I included the cost to purchase, the actual points scored across all of Season Eight, and then the alternate history points scored to get a better sense of the true value of each leader in the opening round games. Again, this analysis necessarily does not include any results from the playoff rounds and therefore undervalues teams of AI favorites like the one put together by antisocialmunky so please do keep that in mind. Nevertheless, I still think that the exercise is interesting as it helps showcase which fantasy teams experienced good luck and bad luck throughout the opening round of games. The biggest and most immediate takeaway from this analysis is that Henrik was absolutely robbed in the fantasy competition. Henrik's expected fantasy points total from the opening round obliterated the rest of the field with no one else even remotely close. Somehow he managed to score zero points from Hannibal, Stalin, Hatshepsut, and Peter combined as compared against more than 200 fantasy points across the alternate history games. Divided out by 20 alternate history games, this suggests that Henrik missed out on roughly 10 fantasy points on average just from the opening round without counting anything that these leaders could have scored in the playoffs. Getting an unexpected 12 points from Augustus Caesar wasn't enough to overcome this putrid luck and I hope that Henrik will consider taking part in the next AI Survivor fantasy competition since this data certainly suggests that he deserves another shot at this contest.

Our actual fantasy winning team fielded by Kjotleik was the best of the rest and stood above the other six competitors, if still well behind the expected fantasy total from Henrik's squad. Kjotleik had three big fantasy scorers in the alternate histories, Justinian and Gilgamesh and Pacal, with two of these three delivering good results in the actual matches on Livestream. The biggest contributor to his winning fantasy team was Gandhi who picked up 7 fantasy points at a bottom-basement spending cost of 15 gold. However, the alternate histories suggest that this was not a very good purchase at all since Gandhi had a paltry 16 fantasy points across all 20 replays of Game Six! You have to get lucky somewhere to win the fantasy contest (I certainly did the year that I won it) and Gandhi's magical run to the championship was the biggest factor working in Kjotleik's favor.

What's fascinating to me is that the next five fantasy teams all scored remarkably similarly across the alternate history games for the opening round, ranging from 234 points for Bellomorphe down to 213 points for antisocialmunky. Bellomorphe had unlucky results from Huayna Capac and Mao Zedong only to see Elizabeth score a highly improbable 13 fantasy points to leave his team 1 point away from taking Kjotleik's crown. J_mie's team tied with Bellomorphe in the actual fantasy standings and then nearly tied again when running this alternate history analysis, finishing only 4 points behind to suggest that their two teams were extremely evenly matched. Jmie had a bad miss on Cyrus only to have this balanced out by an extremely favorable result from Churchill which put this team more or less back where it should have been. Euxaps and Amicalola both did notably worse in the actual fantasy competition while putting together teams that scored nearly identically across the broader alternate histories. Euaxps had the big swing and miss on Julius Caesar and then saw none of his mid-tier selections go on a major scoring run while Amicalola lacked the magic of last year's fantasy triumph this time around. Amicalola was definitely the least lucky competitor after Henrik as his excellent bids on Willem and Roosevelt failed to turn into much of anything while Louis was simply a disappointment. This was a better fantasy team than the 14 points suggested while still being well short of a true contender for Season Eight.

Among the last two squads, antisocialmunky's team is hard to evaluate using this sort of analysis because he was tied so tightly to Mansa Musa. Overbidding for Mansa was all about the expectation that he would win not just the opening round game but also at least one playoff match afterwards - and that did happen! - but this exercise leaves out everything aside from the opening round. Even with Mansa having another awesome season, I think antisocialmunky was a bit on the lucky side to get 7 points from Pericles along with First to Die points from both Charlemagne and Darius. Still, this team nearly won the whole contest and would have done so if Mansa had claimed the championship as he did in most of the alternate histories for that map. Then in conclusion, the clear weakest fantasy team based on the alternate histories was the one assembled by JDP El Grillo. While Grillo's team finished in the middle of the standings based on the Livestream games, this was a fortunate result based on totally improbable outcomes from Brennus and Hammurabi. Grillo's big purchases of Shaka and Genghis Khan turned out to be whiffs in both the Livestream games and the alternate histories which simply put him in too big of a hole to recover from.

Last note: Henrik and El Grillo finished with exactly the same fantasy points based on the official matches that we watched even though Henrik's team was clearly, ummm, a lot better. I think it's a good reason why no one should take these results too seriously - it's all in good fun.

Conclusion

Here's one final table showing the gold efficiency of the auction purchases based on these alternate histories from the opening round. A surprising number of these were pretty close to 1:1 with one gold in the auction converting into one fantasy point across the 20 map replays. I thought that this was a useful snapshot to see which bids were most effective at scoring fantasy points across the alternate histories relative to the cost that the more popular leaders required. Now this is again a limited analysis even aside from not being able to include playoff scoring: it's inherently easier to be more gold efficient when it comes to the cheaper leaders who didn't cost much in the auction. You would much rather have an efficiency rating of 1.25 on Justinian by paying 65 gold for 81 fantasy points as opposed to an efficiency of 1.47 from 17 gold spent translating into 25 points with Bismarck. It's probably more useful just to look at the most desired leaders that cost 40+ gold to see how they stacked up against one another; Mehmed, Hatshepsut, Pacal, and Willem seemed to score the best here with a lot of the most popular leaders having relatively poor value. Again, adding playoff scoring would make this look a bit more reasonable to the favorites if there were a way to do that. (I will note one final time that Henrik had absurdly efficient gold purchases across his whole fantasy team that largely didn't work out due to poor luck.)

Overall, this competition was once again a tremendous amount of fun and helped to make Season Eight of AI Survivor so entertaining to run. I'd like to thank our eight participants in the fantasy contest once again and hope that we can do this again whenever the next season of AI Survivor might be. Kjotleik has an open invitation to defend his fantasy title against what will hopefully be seven new competitors from our AI Survivor community. Until then, thanks once again for reading and following along!