New R2G main page

Dear All,

All R2G’s are now categorized in one of these four groups:

Foundation: These models pick stocks with high liquidity making them relatively easy to trade

Investor: These models offer a balance between liquidity and simulated returns

Trader: These models are the most aggressive ones with either low liquidity stocks, very high simulated returns, or high turnover

ETF: These are ETF models presented as a single group

The criteria for each category is as follows:

ETF: All ETF models
FOUNDATION: (liquidity) * (Ideal#Positions) > $100M AND (Ideal#Positions) > 5
TRADER: (liquidity) * (Ideal#Positions) < $5M OR Alpha > 30 OR turnover > 1500%
INVESTOR: Everything else

We think this new presentation will make it much easier and logical to find a suitable model, and compares apples to apples. Let us know what you think.

Thank You

NOTE: Ideal#Positions is what the models strives for in terms of # holdings, not the current # holdings.

Very nice Marco -

I don’t see Stitts Wealth Creation - S&P 500 / Short S&P 500 in the foundation category. The liquidity is $16.5M and number of stocks is 8.

Steve

Anything with a simulated alpha of 30% or higher goes automatically to TRADER. Both Marc & Paul insisted on this.

While I preferred to be able to see any model on one list I don’t mind the categories.

I find it interesting that Foundation outperformed Trader out of sample.

This is very sensible. Nicely done.

…and it is going to be very interesting in a year or two to be able to compare oos results of trader vs foundation strategies as a whole.

There’s a lot about this that’s going to be very interesting to see.

One might be tempted to look at the categories as being broken down by risk. That’s sort-of the case. But I use the phrase “sort-of” because this is not statistical or academic risk, which, although in use for a long time, has shown itself to be of questionable value in the real world. We strove for something more pragmatic, something that gives a sense of where a model fits along a scale of “How much has to fall into place in order for this strategy to deliver?”

Trader models have the capacity to deliver very powerful returns out of sample. But for that to happen, more has to go right than would be the case for the other categories. Economic and market conditions will likely have to come closer to the average of the 1999-present period than would be the case for other categories, which should be better able to tolerate changing conditions. Ideas will likely have to come to fruition faster since turnover in this group tends to be higher (i.e. it’s not enough to be right; you have to be right more quickly). And with smaller, more focused, portfolios (by design – absent temporary reductions due to hedging or market timing induced cash positions), the tolerance for a bad outcome is diminished. There does appear to be at the start – where we stand today – a large range of out-of-sample difference among the Trader models. So much the better for the designers whose models are delivering. This new interface helps make them more visible to prospective subscribers. And with so many models in this group (100 as of now), this improved visibility should be a very good thing for those designers.

Meanwhile, many investors don’t want to buy into the characteristics of trader. They’d prefer lesser requirements that all go right, even if they have to sacrifice some potential upside. We believe there are a lot of potential subscribers for these categories and we want to make it easier for them to see the kinds of models suitable for them without distraction from models in which they have no interest. We’re also hoping that the more apples-to-apples comparisons will encourage designers (those now present and other who have not yet jumped into R2G) to aim at these categories. Now, they know that their work will not be buried under an avalanche of trader-type models. It will be much easier for those who deliver to these target groups to be seen by prospective subscribers.

Obviously, this is a new endeavor and we all should understand that there will be individual categorizations that are debatable. Hopefully, though, this is a step in the right direction – and we’ll continue watching with analytical eyes.

“Anything with a simulated alpha of 30% or higher goes automatically to TRADER. Both Marc & Paul insisted on this.”

Sounds good.

There is a row titled “Auto-trading via IB”

Do you have any more details on this ?

Auto-trading via Interactive Brokers: we will send the orders automatically to your IB account on rebalance. We’re in the midst of testing and all looks good.

We will only support a few order types initially to keep it simple:

They will all be day orders, and should have 100% fills. We send these orders once to IB and let IB handle it. We will either get a fill, partial fill, or cancel and we won’t retry. We might have few options like a start time, and sells before buys. Other order management strategies will follow that may do changes intra-day.

Initially we will not allow auto-trading in the “Trader” R2G category. We’re taking a cautious, wait and see approach as these models are aggressive. We might allow auto-trading if , for example, the “Trader” R2G has been out-of-sample for 1 year.

And , of course, you will be able to auto-trade your own models - caveat emptor

Very nice Marco!
do you have any information from IB regarding the execution quality of VWAP best-efforts? It would seem reasonable that they would have an idea of average slippage depending on average liquidity or something to have a rough idea of what kind of slippage to expect vs the actual VWAP.

I like the effort to create cumulative stats by ‘model’ type and report that. That’s cool.

However, to make the reported out of sample statistics both fair and meaningful they have to include models that have been delisted - from the time they were posted until the time they were dropped. Not just models currently alive. Do they?

Would also be cool to see rolling correlations between model types. And maybe total number of sub’s per system and or dollars invested by category through autotrading.

That’s not the goal. This is not a “study.” It’s a storefront. We’re showing aggregate characteristics of the products that are available for sale.

Marc,

That’s potentially very misleading to uninformed R2G investors. At the least, I would love to see P123 additionally report what an investor in an equal weighted basket of all of the R2G strat’s from any one category would have earned out of sample. That’s a statistic I have to believe people managing their own money want to see.

You keep saying this is not a study. But anyone managing their own money with R2G’s (and/or trading their money through P123) has to be interested in how this ‘benchmark’ basket is doing, don’t they?

If you don’t want to provide this on your storefront, maybe it can be available as data for the highest end of membership.

Best,
Tom

Like it! :slight_smile:
Andreas

Good development of R2G, in my opinion. I may actually offer some of my models now, as they would fall in the “Investor” category. I previously did not bother because I’ve noticed that higher liquidity models do not seem to attract subscribers, as they can’t compete with “Trader” models on simulated alpha.

[quote]
We will only support a few order types initially to keep it simple:

Market orders can be a problem if a couple of dozen subscribers all send out market orders for the same stock at exactly the same instant.

Relative orders using SMART routing (which is the normal default) can also cause slippage. A relative order means that you are saying: I will always pay more than any other bidder. If another bidder says the same thing then it will start a bidding war. For small orders this is not a big deal; if worst comes to worst you will get filled at the ask price. But for large orders such as when a dozen subscribers all try to buy the same stock at the same time then there may not be enough shares available at the ask price and so the price can just keep moving up and up and up until it gets completely filled at higher and higher prices. This can lead to severe slippage.

FYI: You won’t find this in the IB documentation, but there is a variation of a relative order where you can route relative orders directly to the Island exchange and keep it hidden. This type of order will not move the NBBO. The disadvantage of a Island routed order is that the liquidity available is limited to what is traded on that one exchange.

(Liquidity for Island routed orders doesn’t affect the price but it can be an issue. Contrary to what most people might think, even when a stock is listed on the Nasdaq, many of it’s shares may be bought and sold throughout the day on many other exchanges. For example IB lists 19 exchanges through which shares of AAPL change hands.)

Another problem with relative orders is that either type of relative orders add at least 0.01/share to the cost of the order (assuming penny increments). For low prices stocks this can add up. For example $5,000 worth of shares trading at $2 is 2,500 shares. At a penny per share it adds up to $25 on top of the commission fee.