Thank you both so much! Truly appreciate the "2 cents". First plan is to look into the colocation. We actually have an idea of amount of slippage, so no worries there. And yes, no need to compete with HFTs, just need to get filled at specific prices
If you can get a decent server in NY or Chicago for a few hundred a month and reduce slippage a great deal, why would you not. Lots of great traders start their journey into algo trading that way. Then they realize there are better strategies to be had given the new technical possibilities and then they invest more into infrastructure.
I totally agree that 20-30 ms is going to make the information content (IC) of the data his system receives way way higher than the probably 300+ delayed data he is getting data now. No debate there, I'm just saying there are way more variables in play than just proximity to the host. There's your data providers latency, sterlings latency, who-knows-what-else. Latency itself isn't constant either obviously. It's not link when you ping 8.8.8.8 it says 25 ms 25 ms 25 ms every single time. 500-1000 lots a few times a day isnt in direct competition for HFT profits, but HFT is definitely in competition to be their potential counterparty, right? I'm not saying hosting won't help, I agree with you, it probably will, but I'm just saying slippage is a really dynamic problem and hosting is a step in the right direction, but you need to assess the trade offs of that step (although hosting very cheap for the 500-1000 spy lot crowd!). If you want to compete with HFT, colocation is the cheapest part of the equation and ever aspect needs to work flawlessly. It's not a job for two people. For these guys to decrease slippage my two suggestions would be to 1) get a broker whose order routing provides good fills 2) maybe colocate/host with that broker. Like I said, slippage is a dynamic problem, so although hosting is step one for lowering it, it will still exist, I encourage the OP to host (for a number of reasons outside of just latency), maybe improve brokers, and collect data on slippage to better understand it. Even if hosting doesn't improve his fills, the uptime of hosting will likely pay for itself. Redudancy is critical for automated trading, so you that's a major +1 for your hosting suggestion from me.
But at that level you need your own prime, a few hundred large a month on infrastructure and colo in each major data center. This is not for a guy who is trading via Sterling right now. Otherwise he would not be asking
Haha very true. Like I said, your recommendation is solid. All I add is just that it's a good idea to collect data and try to model out what your E(slippage) will be for each trade! If nothing else, hosting will be a huge plus in terms of uptime.
You're talking about slippage experienced while adding liquidity in a .05%-spread security that trades massive volume. I think that such slippage is inevitable, unless you're one of three parties: 1) a broker/dealer with lots of customer orders to internalize (trade against); 2) a "wholesaler" with enough scale to get a deal to buy customer order flow; or 3) an HFT with a large budget and the know-how to colocate at all the relevant exchanges; in the case of SPY, that also means including the CME at Aurora, IL and the relevant NY-Chicago link. You can mitigate this somewhat, but nowhere near entirely, by following some of the latency-reducing advice that's already been given. Unfortunately, the market is not as straightforward as the NBBO would seem to indicate. It is fragmented, and routing decisions are driven largely by brokers who often put their own interests above the interests of their customers. The HFT and wholesaler business models are very expensive, and that expense shows up in the slippage paid by all the rest of us. On top of that, you also need to consider complex order types, and both how you use them and how other people are using them. If the slippage overwhelms your strategy's profit, you might consider whether it's still profitable when taking liquidity instead.