Design notes — The Alpha Engineer: Transatlantic Small & Micro Cap (Designer Model)

I opened a Designer Model this week. The description provides some details, but a thread gives me more room to explain what it does. There is also a question I would like the forum’s view on. I doubt I am the only one with this issue, but I haven’t found a good answer yet.

Relation to The Alpha Engineer. The model uses The Alpha Engineer’s multi-factor ranking system, the same system I write about on my site. However, the Designer Model runs on a broader universe: about 5,700 stocks versus 3,700 for my own live model. Loosening a few universe settings improved results a little in testing without materially affecting the risk metrics. Also, because a Designer Model might be used by up to 10 people that trade at the same time, its buy rules are much stricter on tradeability.

Universe. Small- and micro-caps across the US, Canada, the UK and continental Europe, restricted to exchanges where Interactive Brokers (IB) is active. Most small-cap models here cover only one side of the Atlantic. My ranking runs across both because my factor logic travels well and the pool of candidates roughly doubles. So I believe it adds something new to the models available here, which some of you may find interesting.

Portfolio rules. Twenty positions, rebalanced weekly, selected from the top of the ranking. Exits follow a single rank-based sell rule. The buy rules do three things:

Buy1 StaleStmt = 0
Buy2 AvgDailyTot(1,0) > 250000 AND AvgDailyTot(1,1) > 250000 AND AvgDailyTot(1,2) > 250000
Buy3 Close(0) > 1

The first keeps out names whose latest financial statements are stale. The second asks for more than 250K USD of daily traded value on each of the last three trading days, rather than as a three-day average, so one busy day cannot carry an illiquid name into the portfolio. The third is a price floor that excludes penny stocks. Annual portfolio turnover runs at about 250%.

My question. I restricted the universe to exchanges where IB is active, so that an IB user would most likely be able to trade every name. In practice, the coverage has gaps: IB lists a country as covered, yet some of its large primary or secondary exchanges stay out of reach. Norway is one example, at least for me as a European account holder. As a result, a few percent of the names cannot be traded on their home exchange. Sometimes you can find them elsewhere; many Norwegian names, for example, also trade in Germany. How do you handle this in your own models? Or should we just accept the leakage? A future update of this model could align the universe more tightly with what IB actually offers. I would rather do that once, and do it well.

Model page: https://www.portfolio123.com/app/r2g/summary?id=1948189
Methodology and research: The Alpha Engineer — Investing with a Quantitative Edge

Disclaimer: this is educational investment research, not personalized investment advice. I may hold positions in names the model holds.

I appreciate the write-up, Peter. Seeing the demo "stock ticker" threads P123 is working on made me think that it would be nice if each DM came with a thread to discuss the model, too. So, love to see yours.

As a DM user, I prefer foreign holdings to be IBKR-tradable

I'm interested in learning more about your (and anyone else's) NOR experience. I tried to buy some positions in the past and never got filled, even after sitting at the ask forever. I recall one position gave a tiny fill, which was worse than getting no fill at all. Then it took forever to offload after I gave up on filling the remainder.

I did some research and found a post where Yuval mentioned he and several friends wrote off trying to trade Norwegian stocks on IBKR for reasons that I believe are similar to what you describe. So I removed them from my process.

More recently, I was experiencing an issue with an ITA stock. I believe the issue with the ITA stock is that the allowable lot-size increment is 1, but the suggested increment is 250. I think any odd-lot order was getting routed somewhere it would never fill. It was interesting because IBKR didn't reject the order or warn me, but once I found the suggested increment and started trading in whole lots, it filled right away (i.e., 5001 shares wouldn't fill at all; 5000 would fill instantly). I’ve experienced this before, but never without getting a rejected order.

Anyway, while investigating that issue with IBKR, I mentioned my NOR problem, and the guy said he used to get calls about it all the time, but he assumed it was no longer an issue because he no longer gets calls about it. (One might think there'd be a more official record, but his anecdotal story sounded promising enough to me) and he mentioned they fixed an exchange-related issue.

He said my problem was that I can't send those orders as hidden. So I tried sending them unhidden, and I was instantly able to fill several NOR stock orders that were previously problematic. I assumed the problem was solved and planned to reintroduce NOR stocks to my process, but now I came across your post.

IBKR also told me that a European user and a US user should have the same experience in terms of which stocks are tradable.

Hmmm!?

The question I have is why for a small/micro cap model the benchmark is set to the S&p 500? There is zero overlap in the between the two universes.

I don’t mind SPY as a familiar benchmark / reference point.

I wish the DM marketplace had a lot more functionality / customization for comparison and also functionality after one subscribes.

To evaluate models, performance-wise, I’ll often add them to a simulated book. From there, you could set your own benchmark if desired. It’s a little cumbersome and more work than should be required, but it helps in evaluating.

The last time I tried buying a NOR stock was two months ago and IBKR informed me that I had no permission to trade but could ask to get permission in a few days. When I got that, I still could not fill the order. For example, GOD:NOR is such a stock I couldn't buy. On the other hand, last year I bought and sold ENH:NOR without issues on that same exchange. I wasn't aware about that hidden/unhidden issue so I will test that next time.

Some ITA orders come in lots indeed. IBKR doesn't give you the lot size so you need to check the exchange to know the lot size. If that information was available, Portfolio123 could fix the increments. The only solution today is to have a size that approximate what the model suggests.

Correct but because I combine all those regions in one model, there is neither an index I can choose from that makes sense. I guess that pro's using this model as part of a book (e.g. for US/EU small cap exposure) use custom benchmarks but I cannot add those. So I am using the S&P 500 because that is what general investors that are not professionals measure themselves against. I use it on my site as well.

Ah, yes, I believe the IBKR-Norway issue you're showing is still ongoing. Some stocks don't have an actual Oslo Børs listing available at IBKR. Some of those have an OMXNO alternative venue.

From what I can tell, GOD appears to be available only on the German exchange at IBKR.

I don't know how to handle that in DM creation, but my platform suggests alternative tickers and, when available, handles the conversion for me with a permanent mapping to something tradable.

I would be just as good with a DM that excludes those names altogether, since I imagine performance would be similar either way, even if I have no way to test that. I'm not sure how possible that is to create.

The problem I had was with stocks that should be quite tradable at IBKR: VEI, OKEA, PNOR, etc. I doubt anyone else is sending these orders hidden, and I wouldn't recommend doing so!

I'm now up to my eyeballs in Norwegian stocks I didn't necessarily intend to own while working through proving I can get fills at size. Hopefully I can get sells at size, too.

About this part, I use the IBKR API and can get minSize, sizeIncrement, and suggestedSizeIncrement.

Some tickers will reject you if you don't size properly, so at least you know. But others will match minSize and sizeIncrement, and if you don't match suggestedSizeIncrement, your entire order appears to be sent somewhere that will never fill.

Some tickers will sit in limbo; others might fill what they can that fits suggestedSizeIncrement and leave the remainder hanging (routed elsewhere, I presume).

Again, I can handle this on my side. But it would all be easier to solve if P123 had API access on its side for DM and Group models.

Since I can handle it on my side, I suppose it could technically be built into the DM as well, but that would require customizing it for a specific broker. While I am admittedly IBKR-biased, I still think that might be worthwhile for others, too, since some DM models choose to hold stocks that I would say most of us likely cannot trade.

Thanks for the insights. Yes GOD is actually only available on the German exchange. I remember having issues with PNOR not so long ago so that was the latest Norwegian stock I tried to trade.

I subscribed to the market data. Apparently, I can trade PNOR now. So the market data needs to be unhidden and it works. Strange. It hasn't been an issue for other exchanges.

I don't know if you necessarily NEED market data (perhaps you do), but without it you'd be trying to place your orders blind and may not be making reasonable offers, which may not get accepted or at least not get filled.

I had market data, was making good offers, but odd things happen when you hide foreign microcap orders. Hidden a little too well, you might say.