Credium Shrimpy Margin `[signal generator] -> [risk allocation] -> [execution]` ## Signal Generator This is where we make predictions. There will be some data that goes into the signal generator and a buy or sell signal pops out of the other side. If you see any bot’s that use “technical indicators” then it’s probably a good idea to try not to make eye contact and just back away slowly. 👀 ## Risk Allocation This takes the buy or sell signal then decides how much to buy. As in, should we allocate our entire capital to this trade or just a portion? Should we buy all in one go or should we average in? So now we know the direction, we know how much we want to buy or sell, next is the part that actually executes the trade. ## Execution See, if you have a lot to buy in one go (say you have to buy $10,000,000 in total for 500 clients) then you probably don’t want to do this all in one trade as you’re unlikely to get a favourable price. 1. **Longevity**. Bots can operate 24/7, humans need to sleep (and rest). 2. **Speed**. Robots operate magnitudes faster than a humans thinking time plus reaction time. 3. **Emotionless**. Robots aren’t driven by greed or fear. They’ll always do what’s statistically more likely to win. 4. **Capacity**. Robots can process gigabytes of data per second. Humans quite simply can’t process that much data in that timeframe. ## Momentum **Prices are up and we think they’ll continue to go up = Buy (or vice versa).** Statistically, most momentum strategies don’t win particularly often but when they do win their gains are fairly large. Win:lose rates of around 55% with gain:loss of around 70% are fairly common with profitable algo’s. ## Mean-Reversion **Prices are up but we think they’re due a pull-back = Sell (or vice versa).** Conversely, most mean reversion strategies win more often than they lose however the gain to loss ratio is smaller. Win:lose rates of around 70% with gain:loss of around 55% are also pretty common with profitable algo’s. A commonly overlooked factor that greatly impacts your profitability is your **fees**. See, Your **transaction costs** (paid to the exchange) and **trading costs** (bid offer) can have a drastic effect on how much your bot makes. **Exchange fees** can vary whether a bot is a liquidity taker (passive) or a liquidity maker (active). The best algorithms will manage their **active to passive trade ratio** and also trade across multiple exchanges dynamically choosing an exchange based on the **optimal transaction costs.** Bot’s that only eke out a small statistical edge can have this completely swallowed up in fees if it’s the aggressor on every trade.