#10: Agreed.
One example: Say your first game is against a non-mover and you win your two next game: score 4+4+4 = 12.
Compare against: one loss and then three wins: 0+2+2+4 = 8. Depending on the number of games in the tournament: pretty big difference.
So we know people late-join to get easier pairings. But in which cases does rocket start fix this problem?
1) In the long tournaments (e.g the hour+ [hyper]bullet ones) the winners score 100+ points. I don't see how the initial streak makes a big difference. Late-joiners will still late join.
2) In the short tournaments where most players play fewer than 10 games players with rocket start will quickly become uncatchable and in some cases tournaments will be decided by "unearned" points.
3) So that leaves us with all the tournaments in the middle. And here rocket start probably does what it is meant to do: Incentivise players to play from the start (without making the first game make or break for the whole tournament).
It seems like it is hard to find a balance by adjusting the points system. It would make more sense (to me) to reduce the effect of players exploiting the pairing algorithm by adjusting the pairing algorithm.
#10 has a nice solution. Simple, easy to explain without looking too artificial, and it's adequately nerfed.
@#11: Back when solutions were being discussed, I suggested modifying the pairing algorithm so that the system tries to pair anyone that just joined and has not been in the tournament previously (during the time that it has been running)* by rating instead of tournament placement. That is, it will wait for one of the most similarly-rated players to be finished before pairing them with the new guy. It seemed like a pretty minor adjustment when I suggested it, but apparently this is difficult because the pairing algorithm is already... large. I took a look in the source code and got completely lost (what the hell is scala?) so I suppose I will trust thibault on this one.
Also, I'm not sure about #11's point 1. In a lot of cases, late joining can basically mean a fast and risk-free 2 points (if there is a lot of variation in strength in the tournament). Would you not take that even if the tournament were long? Even if that 2 points doesn't matter in the end, free is free, and rocket start counters that anywhere. Point 2 seems like a very good one though.
--- wow do I ever look pretentious for using a divider and footnotes ---
*You'd have to check for being in the tournament previously in order to not (rather ironically) make it easier to accomplish the opposite way of subtly cheating: withdrawing and then re-entering right when a player above you finishes a game in order to get a chance to knock them back and take the lead. If you don't check for this, players would be able to very easily choose to get a game with the top guy any time they wanted if they had similar ratings. Apparently players get really crafty when imaginary points or small amounts of money are on the line.
Regarding #11.1, I think I'm a bit confused by what's meant by late-joining: 10 seconds or say 20 minutes (in an 2 hour tournament). Both would affect your pairings, but I see how rocket start reduces the first kind.
#12: Yes the complexity of the pairing algorithm obviously makes it hard to balance. I guess what it boils down to for me, is me not liking how rocket start affects the tournaments I most often play.
Assuming all the drawbacks mentioned, rocket start has obviously some merits. It seems annoying when you see stronger players joining tournaments later come from behind because they face weaker opponents from the start and get quick wins. Whereas strong players staying in the tournament pool from the beginning are matched with opponents of similar skill and hence, reduce their chances of starting a streak. It may have been more fair to give bonus points for the first win for those who join in time, but without streaks.
As for the claims that not everyone has the ability to play the whole tournament, these are funny. What on earth are you guys talking about?.. You can play for as much time as you want unless you are a top place contender, and if you are, I doubt that you face situations when you have a possibility to only play for, say, 18 minutes to win the title instead of 25:)
Sorry for some typos in the previous post.
One more suggestion to admins would be to limit the number of times one can withdraw from the tournament and get back into it. Otherwise this is subject for some fellows to drop out for some time once they get high in the rating along with a streak and logically anticipate to face a tough opponent next.
To execute the latter, there could be an option to withdraw by clicking the white flag button OR if a player decides to just wait in the window with completed game deliberately not returning to the tournament - automatically after some 30-60 secs.