Is the best move to let your time run out?

Already answered that. Chess.com and other interfaces have no algorithm to determine an "unlikely win" which is the same as a "possible mate". Otherwise they wouldn't use any of the material balance rules - except by choice to eliminate unreasonable losses and in contradiction with the FIDE rules. But the real problem is the missing helpmate algorithm.

I understand that.  But this is a little bit different.  In my example, it's the person with pawn who's time runs out.  In your examples, it's the opponent of the pawn whose time runs out.  Clearly a pawn is mating material because it can become a queen.

Actually I understand the FIDE ruling because it applies to OTB play, with arbiters, grandmasters, and etc.  In an online setting with bots, once you open the door to determining whether a position is "winnable" via helpmate, I'm not sure a bot can be expected to make such a determination given the almost limitless variety of pieces and their orientations.

Chess.com's draw (by insufficient mating material) should be changed to include an engine looking for possible ways for a side to win

No, the best move is to resign right before they are about to checkmate you, especially in bullet.

Welcome to the club! Just remember there are many members here who do expect to be the next Fisher and use you as a stepping stone to get there. They care a lot about the rules because there is only one set of rules that will get them to the throne and those are the FIDE rules. They don't mind you will never go there - one competitor down, one billion to go!

Why do you think that because you don't care about the the rules no one can? I want the site that I played over 12000 games on to have the best rules possible. It's not because you don't care that no one can, if you've nothing to add to the conversation the don't bother talking.

Good day.

I know you've answered it, but your answer is dumb. A much better algorithm could be made easily and given that chess.com is the biggest chess site in the world it's not like they don't have the manpower.

The variety isn't limitless though, all that they need is to have a check for single bishops and single knights, most positions are winnable by helpmate (knight vs rook, kinght vs bishop, knight vs pawn and knight vs knight are all winneable just to name a few), they just need to add a few checks for the unwinnable positions.

Martin Stahl confirmed there is no obvious algorithm though there are algorithms which do 99% of the job easily. The problem is that where the exceptions occur and the interface gives the wrong verdict, people will complain not only that they are wronged, but also that they have no clue by which standard - because the decisioning is not transparent. It's of the fuzzy type of "we have done our best and covered most of the terrain e.g. (followed by a list of examples)". Anyway, the algorithm is not as simple and clearcut as the "insufficient material" approaches. That's why it's better when FIDE first authorizes a set of decision rules for computer chess. This gives everybody a point of reference. In human encounters I think it's better to permit arbiters to construct their own helpmate sequences when players won't agree.

Btw, the chess laws have a section with competition rules. This is the perfect place to address such issues to indicate that the principle of dead positions doesn't change but gives way to practical solutions in competitions environments. Composers of chess problems would love that, since they enjoy using the "dead rule" and are not in need of external deciders.

I suppose we disagree about what's a good results. For me an algorithm that in positions where chess.com would give the wrong results gives the right one 99% of the time is a clear improvement. That's just better in every way.

What is diagreeable about having FIDE set the standard instead of negotiating with every site and every chess device about their private approximation?

Because FIDE tournaments have the luxury of arbiters and grandmasters present to determine if a checkmate (via helpmate) can be constructed from a position.  There is not yet a bot that can do this in every case.  And with the millions of games happening on chess.com, assigning a human evaluator in each case that pops up is infeasible.

@Knights_of_Doom: That was precisely my argument! "Setting a standard" means providing a simple (though imperfect) algorithm for all bots to do what human arbiters can do better in OTB tournaments. We had already given up on perfection!

And? We should look for the best option even if it isn't perfect. And one where you draw an obvious mate in one because your opponent times out is clearly way more flawed.

The words "the best option" mean nothing - unless they mean perfection. When you admit you can't get to perfection then you need to specify which standard is acceptable. That standard will consist of "decision rules" and not of "do your best".

I disagree.  Yes getting a draw by running out of time when your opponent has a mate on the next move is bad.  VERY bad.  I agree with that.  However, the aggregate effect of solving that problem by making N vs P a win for N (when P's flag drops) has a much larger negative aggregate effect.  The effect is that all the other N vs P games (which I'm sure number in the millions) will become a race to 50 moves, meaning in most cases whoever has more time wins.  Despite the reality that virtually none of those games are wins for either player.  Having a large number of games become race-to-50 (it's not just N's, it's also B's) is an ugly way to play online, and a heavy price to pay to solve a rare case that happens pretty much never.  Of course it's a matter of opinion.

Best is only ever a relative term. One must compare all the options and see which one is best. The best option will often not be perfect. It's silly to suggest we shouldn't improve the rules because it wouldn't be perfect. Taking something imperfect and making it less imperfect is how things improve.

Do you have trouble with english? Is the concept of best hard to understand?

It's generally extremely to force a trade/capture in pawn vs knight endgames when you have the pawn. Plus with the pawn you can't get mated except in rare cases, so you'd only need 5 seconds to play the 50 moves. So timing out will rarely be an issue. It would only be an "issue" in bullet. Where honestly that's how bullet works, if you're that low on time unless you stop all your opponent's winning chances (three fold repetition or knight capture) you kind of deserve to win.

This is not about the rules or improving the rules, it is about implementing the rules. Nobody disagrees about what the FIDE rules say. There are no predefined options for implementing the rules, there is not a list of choices to pick from. There is an almost infinite scale of approximations of perfection which depend exclusively on how much time and money you want to spend to get higher up that scale. But until you reach perfection there is no "best". It's not a matter of "english", it's a matter of "mathematics". 

Note: "Higher up the scale" means that there are more positions your algorithm can correctly decide as "dead" or "alive". When it can decide every chess position you reached "perfection".

So after all this, the answer to my original question was, yes as of now the best strategy is to let your time run out.  Right?