User blog:BunsenH/Ghazt stats 4

This is the final report on the first phase of our in-depth statistical research into the probabilities of breeding Ghazts from Entbrats and T-Roxes. We've continued with the same protocols as before: the T-Rox and Entbrat both at Level 4 and 0% happiness, with those monsters immediately beside each other. No Wishing Torches were used at any time. All of the trials were run with the Windows version of the game.

Of the 6060 tries, 63 have produced Ghazts. This gives a success probability of (1.04 ± 0.13)%. This is, again, a bit larger than the values at our previous "milestones". The uncertainty continues to decrease, with the larger sample size, but we're getting into the realm of diminishing returns. To get the relative uncertainty down to 10%, with the success probability around 1%, would take a total of around 10,000 trials.

The two sub-groups are the same as before. In the first "minion" group, in which the player level is pushed as high as possible by XP from food production, the levels are now all 22 or 23. This group has 3398 breeding tries and 34 Ghazts bred, a probability of success of (1.00 ± 0.17)%. This is a little higher than the previous value and as expected, the uncertainty is lower. In the other "minion" group, the player levels are kept at 11, the lowest possible for breeding a Ghazt. This group now has 2633 breeding tries and 29 Ghazts bred, giving a success probability of (1.09 ± 0.20)%.

The results from the two groups have been converging, with the high-player-level value decreasing and the low-player-level value increasing. Since the player levels have been diverging, we're pretty sure that player level isn't (directly) a factor in the probability of success. Both groups have been increasing in "wealth", as coins saved, with the high-player-level group accumulating far more than the other. (Those "minions" have also been placing their bred monsters to generate more coins... and to consume all of that food.  The low-player-level "minions" were selling their bred monsters without placing them, to avoid getting the XP.)  Again, given the steadily converging success values, I can't see a plausible relationship between wealth and breeding success. At least in the game. :-)

I think that the original difference was due to an abnormally-large number of successes at the very beginning of the trials for the low-player-level group. The Windows version of the game seems to have occasional periods of high Ethereal success rates, something like the occasional "500% chance of breeding Ethereals" specials that the mobile-app players get... but with no notice from Big Fish. The statistical effect of that original bias has decreased as the sample size has grown.

Beginning in September, we'll be going on to phase 2 of the research, exploring the effects of monster level and monster happiness on breeding probabilities. We'll continue with the Entbrat/T-Rox combination, since it lets us get more trials than other combinations (thanks to that 8-hour breeding time). Half of the "minions" will work with monsters at monster Level 4 and 100% Happiness; the other half will have their monsters at Level 15 but 0% Happiness. Both groups can then be compared with the results from Phase 1. (Of course, it's also possible that breeding success is favoured by having the two monsters at very different Levels or Happiness values, but we think that's less likely.)

The "minions" and I thank you for your interest.

(I've submitted all of our research data to Duhpuck for inclusion in their breeding research project.  This will rather load up their data set with a rather narrowly-biased data group.  If you're willing to commit to a little bit of recording of your results — failures as well as successes, it's very important to keep track of an unbiased data set! — you can help to broaden their data.  There are a lot of variables still to explore, and our own research shows that distributing the research effort pays off enormously.)