Intro

We'll start this blog post with a toy example. If you've played ultimate long enough you've doubtless seen something like it.

You stand on the sidelines of a practice. Your teammate is slapping himself in the face. Not enough to do any damage, just enough to hurt. Dobby-esque, he's punishing himself for throwing a stupid turnover and then getting scored on. He does this every time he makes a mistake.

A coach storms over and grabs his wrist.

“Hey! What did I tell you?”

“Right. Goldfish memory.”

The player puts his hands on his knees and gazes stonily at the field of play, trying to wipe the memory from his mind and concentrate on the following point, which is now ensuing. The coach walks away to bark at someone else.

The coach and the player are both wrong in their approach to the problem. Their main impediment is that they don't understand short-term variance.

Short-term variance

Humans are present-biased. We believe intuitively that the events that happened recently, such as turnovers, have an outsized effect at representing events overall. We have a hard time understanding random chance and see patterns when there are none. We believe if we turn it over multiple times in a practice, we have slipped into a rut and a change is needed.

Numbers help us look at our groups of mistakes with clearer eyes.

If you have a 93% completion rate (the average in the AUDL in 2023 was 92.99%), then you'll turn it over twice in a row in about every 200 passes. You'll turn it over three times out of six one time in every 600 throws. Over a season's worth of practices, you'll certainly throw many more than 600 throws, and therefore more than likely have a time that you turn the disc in over in what feels like a localized, nonrandom event, when in reality, it is just random variance. I can't imagine a teammate of mine throwing three turnovers out of six passes and not feeling like they were in a slump - but they could be playing just as well as when they throw three completions in a row, which for them would occur 80.4% of the time. And this is all with a 93% completion rate, which is higher than many players.

This, in my opinion, is why players so often delcare themselves in a rut, only to find themselves miraculously fixed the next practice. They regress to the mean. Similar to the hothead fallacy in basketball, players feel that they can get “hot” and “cold” by attributing meaning to random occurances. This is how shor-term variance tricks us: we see ghosts when there aren't any.

Goldfish memory

Back to our toy example, this may be what the coach is trying to communicate to a player, in an indirect way. The player's ecent mistakes are not necessarily representative of his skill as a player, so if the player is able to shut the recent turnovers out of his mind, he won't let the turnovers weigh on him. With a “goldfish memory,” the player can forget his recent mistakes, avoid feeling bad, and focus on what he needs to do next.

There are two problems with this:

  1. In reality, it is almost impossible to forget a recent mistake, especially if you are a competitive player. It's really hard to flush that memory out, so most of the time having a “goldfish memory” is impossible.
  2. Mistakes are important in ultimate. In the absence of mistakes, practice is useless - you can't improve without them. Your brain and body naturally feel bad after a mistake because they should: the negative reward is an inborn mechanism that teaches you to behave differently the next time that situation arises. Feeling horrible after a turnover is productive in the same way that feeling great after a success is productive because it reinforces your instinct to do or not do that again [1].

This is why I believe that goldfish memory is a lazy, bandaid fix. It stymies long-term growth and is nigh impossible to attain, anyway. What I would rather have our player do is focus on “elephant memory”.

Elephant memory

By understanding how short-term variance affects our play, we can focus on our long-term memory bank of experiences. I call this “elephant memory.” In my book, this means remembering you past in its entirety (as best you can) rather than forgetting it all. Zooming out allows us to understand our context within our history of successes and failures, which enable us to do two things.

First and foremost, we can understand that a recent mistake is not a meaningful indictment of our ability. Just because you turned it over twice in a row doesn't mean that you suck. We arrive at the same conclusion as “goldfish memory” in this area, which is to not to get hung up on our recent mistake. We can use our elephant memory to remind ourselves that making mistakes, even multiple mistakes in a row, is expected (as proven by our history) and that we are likely to play better next time, as we'll regress to the mean (also proven by our history).

This allows us to evaluate our mistake in context. If there is an identifiable cause of a mistake (or, better yet, a long series of mistakes over a longer time horizon), we can and should feel bad about that specific choice. Equipped with our statistical knowledge, we can learn more efficiently.

Conclusion

Let's rewrite the script on our toy example.

If the coach reminds the player to use his “elephant memory” rather than his “goldfish memory,” he can first evaluate whether there's an identifiable mistake he made, and whether he has a pattern of committing that error. Say the player turned over a crossfield forehand huck over a vert stack to a mid-tier receiver in good conditions against a force forehand after a swing to the backhand side after the mark flared out to stop the around. If the player often makes this turnover, he can and should beat himself up, but will also understand that his playing is effective as a whole - it is unlikely that he will make that same turnover next point. If the player often completes these type of throws, he can rest easy knowing that he will likely play better next time in that spot, as evidenced by his history as a player and his understanding of variance. In either case, he doesn't need to be so down in the dumps that he slaps himself in the face, nor does he need to forget everything.

Let's not forget our recent mistakes by shortening our memory - let's evaluate them in context by expanding it.

[1] You should focus on what you did wrong when you feel bad. You should think “I shouldn't have done X and I won't do it again.” You should never think “I suck as a whole,” because that is unproductive.