We plan on charging a monthly fee, not a transaction fee, so we won't charge for odd lots.

Whatever IB charges a given customer will be defined by that customer's brokerage agreement. In general, their minimum is $1 per trade, so an odd lot would cost $1.

My understanding is that odd lots used to be more problematic back in the day when paper was used for order management. As it's gone digital, no one really cares anymore - it's all just a number. So should you avoid them? They will certainly have higher per-share transaction fees, but obviously not more than a few cents (worst case would be a 1-share order, which would cost $1). I haven't heard that the market price you get is meaningfully different. It will be interesting to see if we can mine that information when we get more orderflow, but for now, my best estimation is that odd lots have no impact on your pricing.

Thanks Dan,

A few more questions:

--Do you have a sense yet (roughly) what you'll charge per month?
--Is there any way, in the backtester, to apply the $1 per trade minimum but use the flat rate of 0.005 per share fee if applicable?

Grant

We have not set a price. We're still kicking around questions like whether it's per person or per algorithm, and whether or not there should be a price difference for larger accounts, or larger number of trades, things like that.

The backtester doesn't handle the per-trade minimums at this point. I think for you now you could eyeball how many trades your algorithm is making. If it's making dozens of 10-share trades, then you should worry about it, but being wrong by a few dollars shouldn't be a big deal.

Thanks Dan,

Ever considered enabling users to code up their own commission routines?

Grant

Hello Dan,

You did mention a figure in the last webinar....

Obviously Quantopian needs to make money unless you are purchased in the relatively near future. Can some of your revenues not come from IB as you will be not only introducing new customers but also faciltating their access to IB? I think any charges should be based primarily on the value of trades placed per month. You could have a minimum charge per month per user running live of, say, $10 pcm.

As you know if you price it wrong you will have very few takers. People will develop algos here and then trade with IBpy from their own applications or from other offerings like PyAlgoTrade. You can of course choose to price yourselves like Rizm where the 'Pro' offering is $249 pcm for unlimited backtests and live trading of 10 algos with 100 SIDs per algo. Good luck with that!

P.

We definitely have thought of custom commissions. I've also been thinking about having your personal commission structure get downloaded from IB - even easier, and very accurate.

I will say, we haven't had as much usage of custom slippage as I thought we would, and that's a much bigger effect on algo performance than commissions are. That factors in my prioritization of custom commissions! It's a sound idea, though, and we'll get to it.

Peter, I have mentioned a number before ;) I'm constantly trying to solicit feedback from people on what our pricing model would be. Sometimes that solicitation takes the form of "here's a number, what do you think?" And sometimes it's "what do you think the number should be?"

I think that in the long run we'll be able to demand some pricing power from IB and other brokers. At the moment, that's not a productive line of discussion for us. Once we demonstrate some volume I think we'll get more traction. It is a dream for us to be able to provide everything for free, and have the broker be the revenue stream. It may be achievable.

Hi Dan,

I tend to think in terms of Vanguard type expenses (fractions of a percent per year). So, if I invest $25K, the annual cost should be in the neighborhood of N*$25, where N <= 10. In other words, Quantopian plus IB costs would be no more than $250 per year. Sound doable?

Grant

I think that other finance vehicles are good to look at for comparison. Vanguard passive funds, though, are likely too low for a good comparison. This is a form of active investment, and the active vehicles are probably better matches. Though I don't think we're going to be able to pull off the "2 and 20" that a hedge fund can ;)

Thanks Jessica,

I figured Vanguard might be an "apples to oranges" comparison, but then again, Quantopian/IB is completely automated, right (i.e. passive)? In the end, it should be practically the cost of electricity to keep servers running (like lots of other Internet-enabled services).

What would be a better benchmark?

Regarding the "2 and 20" since you are just providing the platform and not managing money, I could see the 2% management fee, but you wouldn't be earning the 20%.

Grant

I think the difference between active and passive is more about who is making the decisions. In passive, the decisions are made by the makers of the index. In active, the choices are made by a human. Active only makes sense if you can beat the benchmark. Passive is, definitionally, not going beat the index. Algo trading is about making choices to beat the index (whatever index you're comparing to), and that makes it active investing in my book.

What's a better benchmark? It's a great question, and it's not one that I have a ready answer for. In my experience building companies, pricing and packaging is often one of the hardest part to get right.