Hi everybody,
while looking for advice regarding my modeling/stata problem, I came across this forum. Using the the search function has brought up several interesting threads, allowing me to learn some new aspects about statistics and Stata. However, either my problem has not been discussed or I did not understand someone else's problem being exactly my problem (I fear it might probably be the latter and I apologize for that!). Seeing all the enthusiasm with which the users of this forum give valuable comments for very complex questions, I suppose my question is rather easy for you (unfortunately not for me as I ponder about it some days now). To my data:
I observe data on archery competitions (no team competitions). A match is always between two competitors and divided into several rounds. A competitor must win a pre-determined number of rounds to win the competition. All in all, my dataset contains about 25 000 rounds observed in about 1500 matches. And important information is that there are about 250 competitors in my dataset, however, there are some players that have hundrets of rounds competed in, while others have only competed in one match and have maybe 8 rounds.

My goal is to analyze which factors influence individual performance. My current model includes fixed effects (for competitors) since the skill of the individual will explain a significant part of her or his performance. However, I wonder if can further cluster observations on competitor-match-level? The reason for that is the idea that not only individuals are heterogenous regarding their skill, but they might further not always perform up to their potential. I though about using fixed effects (competitor) AND clustering observations for competitors on the match basis to control for competitors "having a very good/bad day". However, I am not sure, if this is methodically possible/acceptable? My concern is that fixed effects and cluster might interfere? Second, I wonder if the number of observation for some competitors is high enough to use both?

If using both, competitor fixed effects and competitor-match-clusters, is acceptable, what is the stata command to use? I have read so many different possible commands (e.g. reghdfe but others propose "simple" xtreg commands). So far my model has looked like this:

xtreg performance in round j = var1 var 2 var3 ...var 10, fe

(fixed effects are for the respective competitor)

I am very grateful for everybody who read my post until this point. Thank you! And I would be even more grateful for advice. Thank you very, very much.

Yours
Hendrik