User blog:BunsenH/Ghazt Stats Phase 2, Report 3

This is the third and last report on the second phase of our statistical research on the probabilites of successfully breeding a Ghazt from an Entbrat and a T-Rox.

Again, there aren't a lot of differences from the previous report apart from the larger sample size.

Sometimes, a research project gives negative results. This can be disappointing. The promising new drug turns out not to be effective; you find that the rock sample doesn't contain the mineral you were hoping for; you discover that statistically, a lottery is unlikely to give you a secure retirement. Still, it's important to do the research, if only so you can stop wasting time, money, energy, and so forth on dead ends. You can move on to different drug candidates, look for other mine sites, find better ways to invest your money. Negative results are still results.

In this phase of the research, the "minions" and I were trying to find out what effect monster happiness and monster level have on the success rate of breeding a Ghazt from an Entbrat and a T-Rox. They appear to have no significant effect at all.

Of the 5690 trials run in this phase, we got 58 Ghazts, giving a success probability of (1.02 ± 0.13)%. In the group with Level 4 monsters, 100% happy, we had  2838 tries giving 26 Ghazts, a probability of (0.92 ± 0.18)%; in the group with Level 15 monsters, 0% happy, we had 2852 tries giving 32 Ghazts, a probability of (1.12 ± 0.20)%. The differences are not statistically significant. These doesn't seem to be much support for the claim on the MSM Facebook page that "feeding your breeding monsters to higher levels boosts your chances of getting rare monster" [sic].

Combining these results with those of Phase 1 gives a total of 121 Ghazts in 11,750 tries, and an overall probability of (1.030 ± 0.093)%. Notice that even with all of those tries, we've still got a relative uncertainty of 9%. Since the success rate is so low, a single hit affects the results significantly, so getting the uncertainty down takes a lot of tries.

Now, this doesn't prove anything except for the specific case of that breeding combination. It is possible that for other breeding combinations, monster happiness and monster level (and, from Phase 1, player level) may come into play. It is possible that there's an effect if one monster is at a high level and the other isn't, or if one is happy and the other isn't. Or some combination of these, or some more complex interaction with other variables. It's even possible that BFG has decided to mess with the research project: tracked down the games we're using to do the research, and configured their servers to deliver atypical results. It wouldn't be the first research project that got derailed for human reasons, by the researchers themselves or by others. But at some point, one must assume good faith. And, for a project that takes as much time and effort as this, that it's possible to deal with the variables separately.

After some months of determining factors that don't seem to do anything, we're finally on to testing a factor that we're quite sure will make a difference: the Wishing Torches. Each of the "minions" has one torch permanently lit, and we're off and running. We haven't split up into separate groups, for now; we want to get a rough estimate of the effect of one torch, with no confounding variables, as quickly as possible. From the early data, it looks like one torch adds about 1% to the success rate, about what I'd guessed, but there's a huge uncertainty so far.