Of course in positions that the bot doesn't understand, the temperature map is worthless. That hardly needs stating.
I don't disagree that playing out a position enough will eventually give you a good understanding of it. I have personally played out the following position a large number of times and have studied all kinds of interesting decisions that have arisen from it:
| White | 167 |
|---|---|
| |
| Blue | 167 |
| GNUBg Id: 4HPwATDgc/ABMA:cAkAAAAAAAAA XGID=-b----E-C---eE---c-e----B-:0:0:1:00:0:0:0:0:10 | |
But the issue comes down to what the most efficient use of one's study time is. Let's say I have two hours to spend, and two seven-point matches I've just played. Do I take one particular decision and spend the whole two hours playing it out over and over again, generating new blunders in the process to add to my study queue? Or do I go through all the interesting decisions in the two matches and study each one using temperature maps and varying the position and seeing what happens? If that one particular decision is sufficiently important to understand, then it may be worth investing the whole two hours playing it out to get a thorough understanding of it, but for the vast majority of positions, I don't find this to be an efficient study technique.
I'm a big fan of understanding a few positions well rather than understanding a lot of positions weakly, but there are limits. Say I'm given a few months of study time and I have a choice among skimming 20,000 positions superficially, understanding 1,000 positions well, or playing out 50 positions (propositions?) over and over again. I'd go for understanding 1,000 positions well.
Your mileage may vary.