quakers


one half are math professors, the other half is pondering which shoes to wear when they jump off the edge of the earth.

Good read.

At first I was like "yeah, this guy knows nothing about TDM or pickups and forgot that correlation doesn't imply causation, not to mention heavy hitters get weapon drops, obviously they get more frags with LG/RG!"

But then I started skipping through walls of numbers and got to your ending thoughts. Your conclusion was right :)

I ran the old #tdmpickup and one time we (and by "we", I mean wision) ran a bunch of simulations to figure out if our rating formula could be improved based on available statistics. Turned out it could, but the benefit was kind of small :)

That's why we put emphasis on win/loss ratio, along with net and net damage. Especially with bot-made teams, our balancing was trying to bring each game as close to 50/50 as possible, which is an obstacle for what you're doing, but when you rate to balance, then rate based on results, it all magically smooths out by itself :) Legendary tdmers like krysa don't demand everyone to drop them all weapons, and still lead the team to victory consistently. Aimers from CTF can still win just by dealing lots of damage. But hey, that's the beauty of TDM, you can approach it in very different styles on different maps to utilize your players' strengths and there's no winning formula!

If we had way more data from games (weapon control, area control, even silly stuff like leaving ammo for mates or how fast you drop LG for your quad carrier ;)), then maaaaybe you could figure out a formula that makes a good player. But we don't :) Fortunately this is Quake, a stupid 90s shooter, there's no skill in that, only aim, right? :)

That is well written, I like how you linked a bunch of explanatory things for clarity, that really helped.

But to my surprise after filtering pickups you still had 860 TDMers and 500 CTFers left. Lol?

And some of the conclusions are rather obvious, like CTF being FFA-heavy nowadays or like LG+RG being the choice of skilled Jedis, while RL is the unskilled noobtube.
But it is perverted to which extend it applies, if you want to get good at CTF just stop using rockets:



PS: The set of charts with frags = blue (aswell as accuracies = green) would improve in terms of comparability, if all charts in a set would have the same scale of ordinates, but I think it still gets your point across.
Another improvement would be a set of weapon charts, where the ordinates represent damage dealt per weapon, because right now you only have "hits" per weapon which works well for hitscan but in the case of rocket hits could mean anything from 1 ... 100 damage per rocket.
I am not sure what problem you had parsing the formula, but when applications are against me, I usually type complicated formulas in math editors (LaTeX, Word, OpenOffice, Mathematica, ...), take a screenshot of the displayed formula and then include the formula as picture. It is annoying to modify it afterwards and looks like bad resolution when done wrong, but still it is a robust approach when certain applications do not obey.

It is really sad that this thread doesn't get the resonance it deserves, because this is easily one of the most interesting and well done pieces on ESR.

:/