A friend showed me his bot's watchlist last month. 34 coins. He was proud of it. "If one blows up, the other 33 keep me safe."
I asked him what happened in May 2021. He went quiet.
Diversification only works if the bets differ
Here's the catch with crypto. Most altcoins move with Bitcoin, and they move hardest in the same direction at the same time: down. In a sharp sell-off the correlations between coins jump toward one. Your 34 positions turn into one big position wearing 34 costumes.
As far as I know, the research on crypto trend following points the same way. The benefit of adding coins is real at first, then flattens out somewhere around ten to fifteen names. After that you're mostly adding noise. I'd treat the exact number as a rough guide, not a law.
What the extra coins actually cost
Coin number 30 is rarely a liquid market. The spread is wider, the order book is thin, and your stop gets filled further away than your backtest assumed.
Say your fill on a big coin slips 0.05%. On a thin one it might be 0.5%. That's ten times the cost, paid on every entry and every exit, for as long as the bot runs.
Then there's the backtest problem. Smaller coins have shorter histories, and the survivors look better than the average coin did. The ones that went to zero aren't in your data. So the 30th coin looks great on paper for a reason that has nothing to do with edge.
And a quieter cost. Every added coin is another chart you'll feel tempted to override. Thirty-four positions means thirty-four chances a week to meddle.
Why we all keep adding
More tickers feels like more protection. It feels like prudence. I think that's the trap, because the feeling of safety shows up long before any real risk reduction does.
Here's a made-up example to show the shape. Ten liquid coins give you a portfolio that behaves like, say, six independent bets. Add twenty small caps and you get maybe seven. You've tripled your complexity, your costs and your monitoring load for one extra bet's worth of diversification.
I've watched this play out more than once. A coin that traded a few million a day on a calm Tuesday had almost no bids on the worst day of the month. Stops don't fill at the stop price when there's nobody there to buy from you. They fill wherever the book lets them.
Worse, in a crash those small caps can gap through your stop or lose their bids entirely. The protection you thought you bought is the first thing to disappear.
What to do instead
Set a rule for who gets onto the list, and write it down before you look at any chart. Minimum daily volume. Minimum age. A cap on how much of your risk any one coin can take.
If a coin fails the rule, it stays off, however good its curve looks. That's the whole discipline. It's boring on purpose, and boring is what survives a bad month.
Then check how many of your coins really move independently. Take the last year of daily returns and look at the pairwise correlations. If most of them sit above 0.7, count your list as a handful of bets, not thirty.
Ask yourself one thing before you add the next coin: what would this one protect me from that the others don't? If you can't name it, the answer is nothing.
That's how we think about universe selection in our own systems: fewer, more liquid markets, with the rules fixed up front. The methodology and backtest data are at v33systematic.com.