My first naive approach to controlling slippage was to restrict to pairs having sufficient volume. In this respect, Binance is a very appealing exchange, with high volumes. However it turns out that unlike in regulated markets, volume is not a proxy for slippage in cross-exchange comparison, probably because some exchanges fake their volume. On Poloniex, ETH-BTC has about 900BTC 24h volume, while numerous minor coins on Binance are above 1000BTC, with ETH-BTC at 6700BTC. Yet, when you fetch the orderbooks and compare the actual slippage, Binance is not markedly more liquid than Poloniex.
After noticing this, I found this blog post with pretty graphics and that summarizes the problem well.