There are various approaches described in the literature that claim to reliably predict short-term price movements in the range of approx. 1 minute. While the results concerning prediction accuracy look promising, I still wonder how one could leverage such approaches. Background: I've written a simulator which so far implements two basic trading strategies on historical data. The first strategy is straightforward in that it simply market buys/sells an asset when the simulated prediction indicates an upward/downward movement, respectively. Not surprisingly, exchange fees of say 0.1% more than eat up the profits (the relatively high volatility of most coins compared to shares etc. is apparently not sufficient to outweigh the fees for such short time frames). The second strategy I tried is to place limit orders (price = highest bid / lowest ask) to avoid taker fees. However, the probability with which an order gets filled seems to be very low, particularly since the predicted price movement points to the "opposite" direction. As I'm relatively new to algorithmic trading, I googled for market making strategies and studied some papers but didn't really find a solution. Does anybody have similar experiences or hints to related literature? Thanks!