Not to be a buzzkill, but being fooled by randomness is a thing that should at least be mentioned here. Of course Taleb has more to say on the matter in [1].
If statistics showing Messi is a great player were just random we'd expect them to regress towards the mean. It's true that as he ages they have a little, but if we compare him to players who are also his age he remains a far outlier.
And even against the world's best, in the a tournament that matters (ie, this world cup) he ended up being the second highest goal scorer, second highest assist giver, highest chance creator and second highest dribble completer[1].
So yes, I'd like to understand how randomness applies here.
I'm not saying it directly applies here, but it's a caveat worth mentioning.
Consider giving a fair coin to 5 random people and asking them to toss it 10 times. The probability that one of them will get 10 heads is 1 - (1 - (1/2^10)) ^ 5, which is less than 0.5%.
However, if you manage to get 10^2 people to toss the coin, the probability of someone getting 10 heads is about 0.10. This number goes to ~1 with about 10000 coin tossers.
Let's say you manage to get 10k people to somehow toss the coin, and one person, say Mr. A gets 10 heads in a row (as mentioned, the probability of this happening is close to 1).
With enough people tossing coins, you are bound to get a Mr. A. Should Mr. A be hailed as having special skills?
No, not really. Mr. A was just lucky. He was in the right place at the right time. The probability of him getting 10 heads in a row is still only 1/2^10. It just so happened that he was the person who got 10 heads out of the 10k people tossing coins.
Bottom line: With large enough sample sizes, even extremely unlikely events are bound to happen.
Let me clarify again: I'm not saying that Messi is Mr. A!
Yes, of course random walks are a thing. But that doesn't mean that skill isn't also a thing.
In an adversarial, high stakes game where there can be only one winner (like sports, but unlike the stock market where there can be multiple winners) differences in skill show up in statistics well.
There are techniques like true score theory[1] that separate skill from luck by looking at variation (which is essentially a way of looking at reversion to mean as I mentioned earlier).
Sustained performance over time when looking at skill-based statistics in sport (eg, passes completed in soccer) is not something where a random walk applies.
> I'm not saying it directly applies here, but it's a caveat worth mentioning.
No, it's not. It's worth mentioning why it doesn't apply here, but by just mentioning it you imply (incorrectly) that it does, which is both misleading an wrong.
The true score theory is trying to decompose an outcome into true ability and random error. My example was about the random error (which definitely applies to sports, I'm not sure why you think it doesn't). I never ruled out the skill/ability.
It's interesting that you are objecting to my argument by referring to the true score theory, which quite literally provides a caveat to attributing skills to the observed outcomes.
> you imply (incorrectly) that it does
Of course it does because sports outcome are random. Randomness is not the only factor, but it certainly is one.
[1] https://www.amazon.in/Fooled-Randomness-Hidden-Chance-Market...