Player drops: What's the strategy?

I do not have the answer on the player drop strategies. However, here is some preliminary data I got from the current Mystery Tournament through game 90.

Of the 132 managers, I found 8 with a buy-and-hold approach (0 player drops) and their wins to that point ranged from 37-53 (median 45).

18 managers had 1-4 drops, with wins ranging from 34-55 (median 46).

64 had 5-15 drops (34-60 wins; median 46).

42 had 16-68 drops (34-56 wins; median 42).

The above numbers show one can have a winning team in each drop group. How about the different levels of winning? I assigned the managers to the following 3 groups based on games won:

1. 39 managers had 48 or more wins. Their player drops ranged from 0-22 (median 9) and median roster value of $77.34 million.
2. 36 managers had 43-47 wins, player drops from 0-68 (median 10); roster value of $76.64.
3. 57 managers had 32-42 wins, player drops from 0-42 (median 13); median roster value $75.92

Caveat: The above is only a snapshot of the numbers so drawing major conclusions based on incomplete data and small sample size should not be done at home by non-experts. I am not a statistician and I don't play one in the movies. There are lots of games left and many of us are hopeful for a good season with our current or future players!

Very interesting, you might be able to say that the majority of the managers in the last group had slow starts that needed to be addressed. Some may be scrambling to improve right now trying to hit on a great year.

Do the drops include pre-season. Also I wonder what the trends are for the teams in each group. Have they improved in the last 25 games, stayed the same, or collapsed. I don't know if it's possible to check on that. I know I wouldn't.

It is a puzzle with numerous possible factors to consider. The drops were taken from the main team listing, which do not include the preseason. Going back in time would be quite time consuming, but a season end tally would be relatively easy.

In terms of getting player value (and winning games), I think it would be interesting to see how well a theoretical "call your best year" team at a lower roster value would do against the ordinary mix teams at higher values. That seems to be the Holy Grail of the "quick trial, replace, repeat" approach. I am still not convinced 2 innings of relief pitching are sufficient sampling to make a future performance decision. Statistical probabilities anyone?