Quote from: Deprived on June 27, 2013, 09:02:58 PM

Quote from: Bugpowder on June 27, 2013, 08:32:25 PM

Quote from: MPOE-PR on June 27, 2013, 08:18:31 PM

Quote from: Deprived on June 27, 2013, 08:06:43 PM

I considered the other implementation but rejected it as you'd end up with the perverse situation of players running martingales on the investment side where they actually work. That's the problem with allowing minnows to act like big fish - you can't allow them to risk ANY portion of their balance without also allowing them to change it frequently to make it +EV.

I can't see a 'fair' way to allow different degrees of risk whilst preventing people using it to gamble (and in the process removing benefit from all other investors as well as allowing them to sit at the wrong side of the table for what they're doing).

Do go into detail as to this actually working Martingale, as I suspect it's based on a broken model.

The idea probably goes like this :

Hold the majority of coins in balance. Invest epsilon at 100% risk, then 2 epsilon, 4 e, etc if you lose.

HOWEVER, >99.999% of bets on the site are not Max Profit bets. Thus, your investment would not be getting fully at risk each time. Rather, some very small fraction of your epsilon would, except in very rare max bet events. Thus, it would not function as a 'house-advantaged martingale'. Rather it would function as you investing epsilon, and getting 100x * 1% edge * epsilon / house capital = epsilon / house capital returns, rather than 1% * your capitial / house capital = 1% your capital returns.

I am 90% confident that just investing your whole stack at some capital at risk multiplier of say 5x will outperform any house-edge martingale strat.

-Bug

There's two basic models to use for allowing increased risk:

1. The one you describe - where those willing to risk over 1% only get 'extra' action on bets exceeding 1% of all capital.
2. Allocating ALL bets based on ALL risked capital - so those who risk more get a bigger slice all the time.

Perhaps I was not clear. I am assuming model 2.

Quote

You MAY be correct about risking all at 5x outperforming a martingale but it IS the case that there's differences between risking 100% of BR at 5x and risking 5% of BR at 100x. The detail of the difference varies depending on which model you use - certainly in case 1. my instinct is that risking 100% at 5x is far superior due to picking up 20x as much of smaller bets - though you do then stand the risk of losing the lot over a fairly short series of max bets (though for that to be a real risk it involves an idiot gambling on the horribly priced bets where they're odds-on to win but pay through the nose for the privilege - at the 98% level the house edge is effectively 50% : you only make half the profit on winning bets that you would on a fair bet).

The issue here is that you will rarely get close to even 10% action on your invested coins, even at 100% risk level. If selectable risk % is implemented, max bet will go to 1,000BTC. Mean bet size is currently .0035 BTC. So even at 100% risk, only a miniscule amount of your investment will be at risk.

The goal is to maximize TOTAL exposure relative to TOTAL house BR on each bet with minimal risk of ruin. For the first bet, 100% of BR at 5% risk gets you the exact same amount of action as 5% of BR at 100% risk using model 2. There are some interesting differences in variance after bet #1 though. One problem I see is that 5% BR at 100% risk will likely blow that investment portion in a large fraction of max bet events, forcing a manual reinvestment at stochastically determined timepoints. Which would be a PITA.

I'd be happy to continue this discussion in a more private venue. Tipping off people to superior risk and coin-management approaches seems self-defeating.