I believe the first time I had heard of the rounding issue was after the launch of PD3 and one of our developers immediately said that was incorrect. However, I will be having a conversation with our developers and bring up the points you responded with as they do seem valid.
It must be hard having to trust developers, and not really knowing what's right and what isn't. I was never able to delegate anything at JD, because I don't really trust anyone but myself to get things right. I'm not sure if that's a strength or a weakness.
While PD2 was running I was always told/under the impression that PD2 was running at a very very slight disadvantage (1/1M bets). Obviously our intention wasn't to secretly increase our house edge a rather insignificant amount
I think the problem stems from trying to get the rolls to go from 0.00 to 100.00 - it's kind of unnatural to have 10001 different possible outcomes. The new scheme in PD3 is much better, with 10000 possible outcomes (0.00 through 99.99), because they are all equally likely.
In PD2 all the outcomes 0.01 through 99.99 came up one time in 10000, but the highest and the lowest came up one time in 20000 each, and that's where the accidental extra 0.005% house edge came from.
Maybe a simpler example would help. Suppose you want to pick a random number from 0 to 10. You have two 10-sided dice. You can't just roll one of them, because that only picks 10 different numbers, and you need 11. So you roll both of them, giving a number 00 through 99. Then you divide by 10, and round to the nearest whole number.
05 to 14 gives 1 (10% chance)
15 to 24 gives 2 (10% chance)
etc.
but only 00 to 04 gives 0 (5% chance)
and only 95 to 99 gives 10 (5% chance)
So if someone bets the 50% bet, for 1.98x payout: "less than 5"... they win if they roll 00 through 44, and only have a 45% chance of winning, even though half the numbers (0 through 4) are winning numbers for them. So you've advertised a 50% bet but it's really just a 45% bet. It's 5% worse than advertised.
In PD2's case, you're not rolling 10 different numbers, but 10 thousand different numbers, so the error is a thousand times smaller. Instead of a 5% error in house edge, it's only a 0.005% error.
It sounds like an insignificant amount, but 0.005% of 400k BTC (the wagered amount) is 20 BTC. Which is still insignificant compared to turnover, but would probably be significant when shared between the people who lost when they should have won.
Also the PD3 algorithm was fixed right after you noted it to us, we simply forgot to update the verification page.
Fair enough.
It probably seems like I'm shrugging off a lot of your feedback or ignoring it but I've been incredibly busy and haven't gotten the chance to properly respond/look into everything you've suggested. I made sure to get all the feedback I could from you prior to launch and we discussed this feedback heavily with our team. Some of the changes are yet to come after we sort out priorities. While it probably hasn't seemed like it this past week, I've always valued your opinion and input.
I get it. It's obvious you're swamped with problems at the moment. Good luck getting it all sorted out!
