Heavy's Axis Power Craps Forum

Hello:

I love this program and the excellent value it represents. I have “programmed” some simple betting strategies, but I need some advice on making the runs “realistic.”

I know that in order to compare apples to apples, I need to run competing betting strategies against the same roll file. I use my phone number to seed the RNG to get a repeatable sequence of rolls, and by varying the first digit of the phone number from 2 through 9, I have eight repeatable sequences.

Since the advice on money management seems pretty consistent, I have the program “cash in and start over” whenever I’m down to 50% of my bankroll, am over 130% of my bankroll, or experience three no-pay seven outs in a row. I print the results for each of the eight runs and then all runs combined – there’s an example below of the last two entries for a run.

What I’m struggling with is, how many rolls are realistic? I’m currently using 300 (down from 1,000) because I doubt anyone’s going to play for 1000 straight rolls – or even for 300. On the other hand, there should be sufficient rolls to compare the simulated strategies.

How many rolls is enough?

Thanks

//===========================================================================
//Beginning a new 300 roll run with seed 9391192and bankroll of $240
//==========================================================================
WIN GOAL! at 74/74. $316.
QUIT! 3 CO 7s at 120/194. $262.
BROKE! at 105/299. $139.
STOP at 1/300. $153.
//==========================================================================
// $-87 for seed 9391192. -16.392151855635
1 Win Goal(s), 1 Broke(s), and 1 Quit(s). Bankroll $153
Bets Placed: 109. Won: 45. Lost: 64. Net: -19. %: -34.8821548821549
Dollars Bet: $1062. Won: $525. Lost: $612
//==========================================================================
// $-5 for eight seeds combined. -19.2787697156445
10 Win Goal(s), 4 Broke(s), 2 Quits.
Bets Placed: 924. Won: 418. Lost: 506. Net: -88. %: -19.2787697156445
Dollars Bet: $8910. Won: $4795 . Lost : $ 4800

To arrive at your answer ask yourself these questions: 1] Do you play on crowded tables, 2] How fast/slow is the crew where you play, 3] Are there typically alot of prop bets where you play, and 4] How long do you play.

Ideally you're asking the first 3 questions to determine the number of decisions per hour [rolls] and the fourth to get your multiplier. I have seen tables that slowed to a crawl and did maybe 25 - 30 rolls per hour due to the first 3, I have seen tables that likely were around 120 rolls per hour or slightly higher. Obviously a game on a 25-30 roll per hour table is painfully slow so odds are if you didn't walk up to a hot roll you wouldn't stay... for the faster table 300 rolls is roughly 3 hrs and I can see it happening. The issue I see is everything also depends if you are looking to see a per session average vs a broader "how good is this strategy" criteria. The smaller the roll sampling the more likely you'll see variance whispering things in your ear that you shouldn't listen to. The larger the samplings the more you'd see how the strategy works once it got outside those quirky hot or cold hands from a session. For me, I would prefer the longer roll criteria since I understand that all things are possible... its just that most are improbable, and I would need to determine how probable vs improbable a given result should be long term.

Thanks, guys. You've helped me see that I am confuisng two different things. Me buying in and playing for an hour or so versus the long term viability of the strategy I'm testing.

Thanks again!

twizforpros wrote:Thanks, guys. You've helped me see that I am confuisng two different things. Me buying in and playing for an hour or so versus the long term viability of the strategy I'm testing.

Thanks again!

Exactly. I'll give you a hypothetical that should illustrate why this is such a bad idea to go forth with. Lets say I handed you $100 to bet $1 at a time on a 12 and on a hit told you to parley... now lets say that night fortune smiled brightly at you and about 20 rolls in the drunk guy that had been such a pain all night rolls 3 12's in a row. On the face of it you've stumbled on an awesome system to make a great deal of cash... you immediately color up and rush to quit your job and see your real estate agent so you can move to Vegas to be rich because lo and behold you turned $100 into $1,990... great, until you do the math and find out long term if you walked in with $10,000 and did this statistically 1x out of every 1,296 rolls should return your parley meaning for every nearly $1,300 you'd get $930 of it back and 1x in 46,656 rolls you'd get those 3 hits in a row.

The problem is that you can and will see really weird things happen on a table short term, 1 night I tossed 7 hard 4s in a hand before I brought back a 1-3 instead of a 2-2... great, bet the entire farm on me doing it again... except I have had several hands with eight 4s but not a single one other than that one had more than maybe 2 or 3 hard 4s in it let alone in a row.

I once told someone that statistics is truth in large numbers and variance is lies in small ones. Anytime you want to know the truth in numbers you should look for the large not the small ones, 1 roll tells me nothing, 1,000,000 tells me everything I might ever want to know as long as I am looking at general trends rather than specific patterns, etc [if I am looking for my signature number for instance instead of looking to see if I toss a 6 after tossing a hard 4 or 12, etc].

One other thing to remember is that even if you ran your tests out to 20,000 rolls or even 2 million rolls you'd likely hit either your loss limit or win goal long before that if they were reasonable or your strategy was decent... but what you want to know with purely random numbers is how this thing plays out across the entire time you use it.