For those readers who can’t wait for my reply to Vladimir: Vladimir’s comments are currently being discussed on Chess Chat.

]]>Being a chess player (an IM) and an academic with 15+ years experience (Economics, Econometrics) allows to me to look at this analysis differently. Unfortunately, this is a back of the envelope calculation and most likely, it is incorrect one.

1) Econometricians are using notions of confidence intervals. So the goal is to come up with the probability that you can reject the null hypothesis with. Calculating a specific number has very little meaning in the scientific world.

2) True scientists are not meant to be emotional. For example, the claim “Conclusion: Mihaela Sandu is innocent” is non-scientific. Instead one could say – The claim that Sandu was cheating can be rejected with probability X.

3) You are not supposed to make assumptions like “It seems reasonable to assume … and therefore I have assumed a drawing probability of 0.30.” Instead, you are supposed to derive a distribution of drawning chances and incorporate it in the analysis.

4) Finally, the derivation of 0.3 probability makes no sense to me. Your earlier calculations were produced under different hypotheses. One can’t just combine all these numbers and plug them all into one formula.

I do not want to sound negative. However, it is extremely sad to observe somebody making so strong claims with no real scientific backing.

]]>ad 1)Most of us would be severely disturbed in our performance under a cheating accusation, just like other external factors (our health, bad news from home etc.) would affect our play. But I am not sure what we can conclude from this with regard to a confidence/boosting effect and the size of it. This is tricky territory indeed.

ad 2)If cheating is taking place and all of a sudden the cheating mechanism is not functioning, I would doubt the cheating player (under the pressure of an ongoing game) would apply a different cheating mechanism right away. It would be more practical for the cheater to accept being on his/her own and play their best under the circumstances – i.e. at their “honest” performance levels. In this case I assumed cheating has a big effect, as this is the assumption putting the innocence hypothesis under maximal pressure. With a lesser effect of cheating on performance level, the probability of innocence given evidence is even higher.

ad 3)The burden of proof should definitely be on the shoulders of the petitioners, not anywhere else!

ad 4)I hope they will!

]]>1) I think the effect of self-confidence could have been put even higher taking into account that she lost all games after the open letters came out. Thus, Sandu’s play is likely to be affected by external factors — perhaps more likely than average. We’re side-stepping to psychology here, so it might be tricky.

2) The actual difference between, and the different effects on a person’s strength between cheating by using the online transmission and cheating “on your own” is highly unclear. In some cases the cheating becomes completely impossible, sometimes it only makes it slightly more difficult, and in other cases it doesn’t matter. This is mentioned when you speak about the difference of the scientific approach and the courtroom approach. Since it is highly unclear to what extent the absence of live transmission drops the strength of a cheater, in my opinion it’s hardly possible to say anything for here (but I understand that you had to make choice because you were making a point).

3) In my news report I have refrained from analyzing games to make a point, as I felt that would approve shifting the burden of proof. Your article seems to support my decision!

4) If anyone should read this piece, it’s the ladies who signed the open letters.

]]>Anyway, I think it would be easier to understand if one does not mention the probability of infant homocide. After all, that is also a conditional probability, namely the probability of homocide being the reason of an infant’s death. Instead, simply presenting the compound probability of infant death would put Meadow’s 1/8500 in immediate perspective (especially as we assume there are only two possible causes of the death). In fact, if one did the same for the case of two infant deaths (instead of simply squaring the one death figure which is of course dubious), one would account for the genetical and environmental issues of the footnote. I’m not saying your reasoning is wrong, just a little unnecessarily complex.

]]>