[Warning, long post, TL;DR included.] So, I'm sure we've all had this happen to us; going an entire map innocent, then getting T 6 rounds in a row. To me, that's a little unfair, but it is what happens sometimes when the system randomly selects the next T. What I want to possibly see implemented is a Dota2-esque PRD (Pseudo-random distribution) for T selection. For those not familiar with PRD, it is used in Dota2 when if, for example, an attack should proc a critical. [Warning, probability notation ahead] P(N) = C * N, where C is a constant derived from the expected crit chance (which serves as the initial chance and the increment for every N), and N is the number of attacks since the last successful proc. Therefore, as you don't proc a crit multiple times in a row, the chance to proc one goes up each time it doesn't proc. Make sense? Using this approach verbatim wouldn't work on a TTT server. It'd be pretty easy to figure out who would get T next given X person didn't get T a certain number of times in a row. The main change to the system that DOTA2 uses would be that C would be chosen from a range of values (which would have to be tuned through testing). In theory this sounds all good and dandy, but it will have to be heavily tweaked, in addition to the fact that this will be more pre-round shit that needs to be calculated. So if you give me the goahead, I would need the way current Ts are selected and then I'd derive the constant, C, from that. If it's just 1/X number of players, then that's simple. From there, I'd run simulations and see the difference between who gets T round with the current system and with my new system, and then continue to tweak the code until runtime is acceptable and P(Traitor|N) [Probability of getting traitor given N previous Inno rounds] is acceptable. TL;DR I want to create a better system for selecting traitors that'll spread out who gets T. Multiple occurrences of T/Inno will still happen, but will be less frequent than the current system. Main downside is computation time.
+2130219318 I really like the idea of this.. It will stop all those people from complaining about how they haven't got T round in "LIKE 7 MAPS DUDE HOLY CRAP WTF". Although i will miss getting traitor 6 times in a row. I feel 6 is the highest, since that's what most people report as their highest. I wonder if anyone has gotten more than 6 before?
I have gotten 8 in a row without buying it. Magma new version all 6 rounds and then 2 rounds into Clue.
I was actually the one who told him to do this *proudness* lol. But yea the whole proc idea, like Dota 2's would be great. I always have players complaining about their odds of being Traitor. +1 NOTE: It gets me sad how Event Horizon is Traitor almost every round. XD
Alright, I'll start prototyping the system with some trials (aka, Monte Carlo simulations) and post the results. For the "results", I'm thinking of posting the following. 10 rounds x 3 (different maps, with + w/o carryover of N) of T selection 15 players - 25 players - 36 players Which will include... Each player's probability of getting chosen as T for each round If they got T Longest T streak Longest inno streak This is definitely going to be a shitload of data, so it'd be nice if I could get an admin to be on board with this so I can get feedback on the results. (it'll take 10 minutes tops, I'm not going to bore you with stupid shit). I'll post back tomorrow with the results.
It wouldn't hurt to start prototyping, but I would really get Highwon's approval for something like this.
Yup, I plan to! Results will give me something concrete to show so when I say my system is better, I have a reason to think so rather than "yada yada theory bullshit."
I had very bad luck of getting T rounds when I was a regular player, then as soon as I became a staff member I started getting traitor all the time... coincidence, I think not! But in all seriousness, I like the idea you have presented, so +1
