altf4 / omniscience Goto Github PK

Are you a designer? Omniscience really could use an icon for the app. Something perhaps inspired (but not exactly, for copyright reasons) the card Omniscience?

Thanks!

I do not have an iPhone. It would be helpful if anyone who's a developer could test run Omniscience on an iPhone.

Currently, it treats it like it's still the last round

In the final round of swiss, players are sometimes put into a prisoner's dilemma. A classic example is when 6 players with the same match points are competing for 4 open slots.

Each of the players realize that they can't all draw in, so they play. Two of the matches finish. The last two players are now able to draw and both get in. Thus all players are incentivized to finish last.

Or one of the matches can immediately and vocally declare that they are drawing before anyone plays. Thus forcing (or perhaps daring) the other two matches to play it out.

Obviously this is a scenario where monte carlo simulation does not adequately model the real world dynamics. So it might be useful simply to notify a player that they're in such a scenario. But how do we do it generally?

One assumption that Omniscience needs to make is what the rate of unintentional draws is. This is currently hardcoded to 1%. But it surely varies from format-to-format. And possibly by event-to-event. Maybe it would be better to derive this value empirically from the tournament itself.

On the last round of an event, user should be able to:

1. Click on a user in the standings.
2. Select Win/Lose/Draw for that user (and necessarily the inverse for their opponent)
3. Run a new batch of simulations, with the given user-supplied results baked in

Can we, given the last round standings, derive a "worst case scenario" set of results?

It would be something like...

• All of our previous opponents lose (making our OMW and OGW percents go down maximally)
• We lose the maximum amount of games for the simulated result. (0-2 losses and 2-1 wins)
• For each match capable of tying / exceeding our match score, the player with the best breakers wins.
• For each match capable of exceeding our match score for both players with a draw, they choose to draw.

I'm not 100% sure that this is guaranteed to be fully actually the worse-case. Some of these stipulations can conflict with each other. But we can't simulate every possible way to run a worse-case scenario.

But at least it ought to get close to worst-case. By throwing this scenario in there, it ought to give us a lot more confidence in a "100%" result. If all of our ordinary simulations come up as positive, it's probably worth-while to get a sense of whether you're "actually" locked.

There's quite a big difference between 100% and 99.9% since we can't guarantee that the worst-case scenario is ACTUALLY rare. Some players may choose to play it out if they know for sure that there's a series of events that cause them to miss top 8 with a draw.