Project Stake: Dice Gameplay Integrity Analysis – Gold Coin vs. Real Money vs. Simulated Play
Happy New Year Everyone! Not much play over the holidays, but we were able to run a comparison study. Let’s dive in!
If you remember from the last post I introduced a “Winning System” devised to evaluate the randomness of Stake’s Originals. For clarification this system can be used for every day play, but I would proceed with caution as Martingale type systems can destroy a bankroll in a hurry when you least expect it. The system was created primarily to create a statistical approach to evaluate patterns outside of just individual bet win chance. By breaking the experiments into trials based on hitting a defined profit or loss stop you can start to see how Stake’s programming works when dealing with increased bet sizes..
In the experiments for validation of Stake I stopped play if Level 5 was exceeded, because why throw away money if the game is rigged. But by breaking my bets into trials you can get a picture for what is going on with Stake’s betting algorithm. In the end statistics should not lie, right? I have recapped the betting system and trial levels below. Take a look for yourself and see, are the outcome of Stake Originals truly random? After running trials with real money, Gold Coins, and simulated numbers, I’ve come across findings that are hard to ignore.
The Experiment
I conducted three sets of tests using Stake’s Originals “Dice”:
The “Winning System” used for validation:
- Win Chance = 40%
- Starting Bet Size = $.01
- Increase Bet Size After Winning Bet = 30%
- Decrease Bet Size After 2 Wins in a row = 50%
- Profit Stop = $.08
- Loss Stop = $.40
- Level Escalation = 3x (When advancing Levels, increase current starting bet, profit stop, loss stop by this number)
- After Winning 2 Levels in a row reset to Level 1
- After Winning at Levels 4 and 5 reset to Level 1 (to reduce risk)
- Stop Play after failing Level 5
- Real Money Trials:
- 5 sets of 100 trials each.
- Starting bet: $0.01
- Progression: Bet size increased by 30% after each win.
- Profit Stop: $.08
- Loss Stop: $.40
- Total Bets: 19,774
- Bets Won: 7,869
- Bets Lost: 11,905
- Bet Win Rate: 39.79%
- Total # of Trials: 494
- Trials Won: 358
- Trials Lost: 136
- Trial Win Rate: 72.47%
- Total Profit: -$66.02
- Level 5 Failures: 5
- Only 1 trial set was profitable (+$7.98).
- These values do not include rakeback or bonus money.
- Gold Coins Trials:
- 5 sets of 100 trials each
- starting bet: 100 GC
- Progression: Bet size increases by 30% after each win.
- Profit Stop: 800
- Loss Stop: 4000
Summary of Gold Coins Trials:
- Total Bets: 16,916
- Bets Won: 6,805
- Bets Lost: 10,111
- Bet Win Rate: 40.23%
- Total # of Trials: 500
- Trials Won: 407
- Trials Lost: 93
- Trial Win Rate: 81.40%
- Total Profit (if converted to same ratio as Real Money Experiment): $18.23
- Level 5 Failures: 0
- 4 out of 5 trials were profitable. The final trial had a loss of $3.28.
- Python Simulations:
- Ran over 50,000 trials using a custom script built in Python to mimic the Stake’s Originals game using a random number generator with a win chance setting of 40%.
- Replicated the same conditions to compare results against Stake Originals.
- Total Bets: 1,718,784
- Bets Won: 686,957
- Bet Win Rate: 39.97%
- Total # of Trials: 50,000
- Trials Won: 39,669
- Trial Win Rate: 79.34%
- Level 5 Failures: 19
- The simulations experienced a total of 19 level 5 failures, averaging 3.8 level 5 failures per 10,000 trials.
- Level 5 was reached about 17 times per 10,000 trials, suggesting we could expect to reach level 5 approximately 17% of the time in a set of 100 trials. So 4 out of 5 sets of 100 trials should not run into Level 5.
- A noteworthy observation from the real money trials was the trial-level advancement and how Stake handles increasing bet sizes while maintaining a 40% win chance.
- In the fifth set of real money trials, play was stopped at 94 trials because 5 consecutive losses occurred.
- While such a streak can and will happen, mathematically it should occur approximately 4 times in 10,000 trials.
The Results
| Real Money Trials | Gold Coins Trials | Python Simulations | |
|---|---|---|---|
| Trials Completed | 494 | 500 | 50,000 |
| Bets Completed | 19774 | 16916 | 1718784 |
| Trial Win Rate (%) | 72.47% | 81.40% | 79.34% |
| Individual Bet Win Rate (%) | 39.79% | 40.23% | 39.97% |
| Profit/Loss | -$66.02 | +$18.23 | N/A |
| Level 5 Wins | 2 | 1 | 85 |
| Level 5 Failures | 1 | 0 | 19 |
Observations
- Win Rates:
- The win rate for Gold Coins and Python simulations aligned very closely with expectations.
- Real money trials, however, showed a significant drop in trial win rate and a slightly lower individual bet win rate.
- Profitability:
- Despite the same betting strategy, real money trials resulted in losses in all but 1 trial, whereas Gold Coin play was the exact opposite.
- Behavioral Consistency:
- Gold Coins and Python simulations behaved predictably. In contrast, real money results seemed to deviate in ways that suggest potential manipulation or adjustments.
- Level 5 was encountered 3 times within 500 trials in real money play. That is occurring at a frequency 3.6x higher than simulation.
- Trial-Level Analysis:
- The real money trials’ frequency of extended loss streaks during trial advancement suggests potential discrepancies when compared with simulated outcomes.
- Simulations show that level 5 failures are rare, expected to occur only 3.8% of the time for every batch of 100 trials.
- Level 5 occurrences in simulations suggest it would be reached approximately 17% of the time in a batch of 100 trials, which was in line with the Gold Coin play.
Why This Matters
Stake Originals markets its games as provably fair. If the outcomes differ significantly based on the currency used, it raises questions about transparency and fairness. My results suggest that the outcomes with real money are influenced by factors beyond the game’s supposed randomness.
Is this enough data to say Stake Original’s is not playing fair?
If you’ve experienced something similar or want me to investigate a different system feel free to reach out. I look forward to hearing your thoughts? See you soon.
