Portfolio and book performance

Research
1

I have a question on how portfolios and books work together.

Let’s say I have 2 portfolios, Port1 and Port2, which I put into Book1 at a 50%/50% ratio, rebalanced annually. When I look at the individual portfolio transactions, I see buys and sells every week. However, when I look at the book transactions, I only see the annual rebalance transactions.

Does the Book1 performance show:
A) The underlying performance of Port1 and Port 2, rebalanced annually to get them back to a 50/50 split; but the book omits the underlying details, so I need to look at the individual ports to see the weekly changes, or
B) Book1 freezes Port1 and Port2 and holds the initial positions for one year, before updating at the annual rebalance?

I suspect the answer is A, but I would appreciate it if someone could verify that.

If the answer is A, does that mean I have to turn Port1 and Port2 into live ports, and then create a live book using the two live ports?

2

Right, simulated books just trade the equity curves of the underlying ports, not the stocks. So there is no detail to see.

Live books take buy and sell signals from the underlying ports and then trade those stocks in the book. If you’re using TRADE, you’re initiating the trades from the book, not the ports. The book must rebalance every time the ports rebalance so that positions can be updated. However, the book may rebalance assets less frequently (this is where terminology between rebalancing vs reconstituting gets confusing).

3

So the first thing that you need to know is that there are two sources of transactions in a live portfolio: Changes as a result of changes in the composition of the underlying portfolio and changes generated by the book itself to bring the proportions of the assets to an acceptable allocation.

The second thing is that, generally speaking, no system on Portfolio123 does anything when it’s not time for a rebalance. (Sole exception: The stop-loss tab in a simulated portfolio.)

If you have underlying portfolios that are rebalancing, say, weekly, and your book is rebalancing annually then you won’t get the changes to the underlying portfolios at all in a live book except for once a year when the book will both apply the changes that have happened to the portfolio over the past year and then bring them back to their target allocation.

This leads to the conclusion that your live books should rebalance at least as often as your fastest rebalancing asset. (Or at least that you should hold to a period that you’d be OK with not applying the transactions that occur during that period.)

By the way, that’s not what’s happening in a simulation book. A simulated book is looking at the underlying portfolios as if they were mutual funds. It’s only taking the equity curves and then applying the weighting to them. This implies that you’re getting all of the interim transactions even though there’s only the few transactions to rebalance.

4

So, if I understand this correctly, when I run a sim with 2 ports that have low correlation, the book performance will show me some of the benefits of low correlated assets such as a drop in volatility and improved risk-adjusted returns. If I have realistic assumptions for the underlying strategies, commissions and slippage, haven’t curve fitted, etc., etc., then the resulting book performance may be indicative of future performance. If I run a Monte Carlo analysis on the expected returns and volatility, I should have a pretty good idea of what to expect on a go forward basis.

However, this is not the case with a live book set to annual rebalance, since the live book is not picking up the weekly transactions. How do I set up the live book in order to 1) capture all of the underlying transactions of the ports on a weekly basis and 2) capture the annual rebalance to get the ports back to the initial weightings?

5

The easiest way is to set the book to rebalance weekly. The main purpose of the live book is to put combined strategies into action easier. The standard way of preventing frequent rebalancing transactions to get back to a target weighting is to use the Minimum Transaction Size setting. You will never prevent changes due to changes in underlying holdings, so you’ll only prevent reweighting transactions, if that makes sense.

6

So, if I have a $100k book with 2 ports that are equally weighted (50/50), set the Rebalance Frequency to Every Week and the Minimum Rebalance Transaction to $1mm, I will have created a reconstitution tab rather than a rebalancing tab. That is because the book will look at the ports on a weekly basis and mimic the underlying portfolios, but because the minimum rebalance is $1mm, it will allow the initial 50/50 weighting to change.

If Port1 doubles after one year, while Port2 remains at $50k, the weighting at the end of the first year will be 66/33. How do I get it back to 50/50?

7

Manually change you min book rebalance level to 0 (or a small number) that one time per year, I think…
Then set it back to your $1mm for the next 12 months

Jmh

8

Jerome and Ihor (racerguy),

The Book needs to have weekly Rebalance, and not annual if the assets (ports) you’re using have weekly rebalance. Bottom line: always set your Book Rebalance the same as your Simulation Rebalance. During this, your Book will always reflect the changes to positions (reconstitution) and changes to the weights (rebalance) that take place in the underlying Portfolios/Sims. Your Book is supposed to keep the two assets in the proper weight relationship, i.e., 60%-40% or 50%-50%, etc. weight for your two assets (or three, four, etc.).

In BOOKS, you want to select rebalance settings that will keep the relationship between your contributing Sims close to your target relational weight – but you also want to minimize unnecessarily small transactions that can run up commissions and reduce returns. To do this, the Minimum Rebalance Transaction (MRT) is the setting you will adjust. Unfortunately, the Book Simulation and the Live Books are inconsistent in the way they handle transactions for relational weighting.

For example:

In the Book Simulations, under the Rebalance tab, you have two possible settings: 1) Minimum Rebalance Transaction (MRT), and 2) Asset Tolerance – and they are both in a Percentage amount.

[color=royalblue]BOOK SIMULATION - REBALANCE SETTINGS[/color]

In the Live Books, under the Rebalance tab, you are only provided with Minimum Rebalance Transaction (the Asset Tolerance disappears) – and the MRT it is in a Dollar amount only.

[color=royalblue]LIVE BOOK - REBALANCE SETTINGS[/color]

A percentage amount across the board is more logical to me because I don’t want to be continually adjusting the dollar amount as my Live Book gains value every week, month, and year (at least that’s the plan). Having to remember to check this every so often is just one more thing to do. I have settled on putting a tickler on my calendar to prompt me to adjust the dollar amount in my Live Books every quarter as they increase in value.

In the Book Simulations, I want to minimize excessive small, costly transactions, but I also want to keep the weights near my optimum to get the highest performance. I have settled on using a Minimum Rebalance Transaction (MRT) of 5% and an Asset Tolerance set at 5%. I find that if it is set higher than this, they can really get out of whack.

Some Books show a higher AR in the Simulations if I use a higher MRT, such as 20%. But a higher setting for MRT can cause an Equity/Fixed Income Book to get really out of whack. For example, a 60-40 weighting scheme can reach an 80% weighting for the Equity portion, and only 20% for the Fixed Income portion as the equity asset significantly outperforms the fixed income asset! A 60-40 target can become 80-20 if you don’t stay on top of it. For this reason, I have settled on 5% for this setting.

Now, it is a mystery to me what happens to the “Asset Tolerance” setting when a Book Simulation is converted to a Live Book! Obviously, it disappears, but how is that handled by the algorithms? What should the Asset Tolerance be set at in my Book Simulations to duplicate its disappearance when my Book goes live? I have yet to find anyone who can explain this to me, but perhaps someone will now (hope).

Also, the other big mystery to me is why we have a percentage for these settings in the Book Sim and a dollar amount in the Live Book. As it stands now, we cannot accurately simulate the performance of any Live Book. It kind of feels like the Book project is unfinished, but maybe I’m just misinformed.

Perhaps Paul, Aaron, or Marco or another p123 member can address these mysteries and complete my explanation. I would love to see that explanation also!

Chris

9

The live book and the simulated book work entirely differently.

The best way to think of a simulated book is that you are simulating investing in two or more mutual funds with each of those mutual funds based on your portfolios (or a public portfolio or Designer Model, or a specific ticker). That’s just what it’s doing, really: How would ABCDE and FGHIJK have done in a 50/50 mix over the length of the book simulation. It’s only looking at the equity curves.

A book simulation knows nothing of the underlying holdings. That’s why it’s using a percentage to try and hold down the number of transactions.

Another major consideration/stumbling block with book simulations is that, because it knows nothing of the holdings it’s very possible that your returns were a result of being very overweighted in a single underlying security. If your two portfolios are, together, 95% in AAPL then you’ll probably have done really well over the last decade without really learning anything.

A live portfolio is looking at the current holdings of each portfolio and then figuring out the proportion that it should be holding for a combination of portfolios at the indicated weights.

So imagine that you have two portfolios that both invest in ABC, one at 25% and one at 50%. You tell the book to combine those two portfolios at 50/50 and invest $100,000. It will figure out that it should have (.25*.5)+(.5*.5)=.38, or $38,000 in ABC, and it will generate a suggested trade for that amount.

In most respects, then, a live book is going to function as a portfolio made up of several live portfolios instead of the mutual-fund-like behavior of the simulated book. And because the two varieties of books are so different under the hood, they have different settings.

Given that the live portfolio deals in dollar amounts, it’s usually easier to just use that to eliminate the little transactions that the live book throws out. The live book is VERY prone to minuscule transactions. It will easily throw out a bunch of $10 trades to get everything juuuuuuuust right if you let it.

10

Paul,

But “little transactions” is relative and varies, depending on the total size of a portfolio. For example, if I want to stick with the optimum minimum of 5% for transactions in a one-million-dollar portfolio that would be $50,000. But for a $100 million portfolio, 5% transactions would be $5,000,000. As the portfolio progresses in size, that amount has to continuously, manually, be adjusted upward. Why not at least offer us the ability to stick with percentages for our minimum transactions in the Live Book?

Also, why make us convert percentages in the Book Simulation to arbitrary dollar amounts in the Live Book? I can test to find the optimum percentage-size minimum transaction in my Book Simulation, but then that information is tossed out the window when we convert to a Live Book.

I understand your analogy to mutual funds in the Book Sims, but don’t see why Live Books are so different from Book Sims in this Rebalance component.

I don’t get it… (But then, there are a lot of things with that status, haha.)

Also, what happened to “Asset Tolerance” in the Live Books after we are allowed to find the optimum setting in the Book Simulations? It disappears. As I said, it feels like the Books feature didn’t quite get finished. But maybe I’m missing something.

Chris

11

Jerome,
Thanks, that is a good suggestion.

Chris,
You raise some good points. That is why I am trying to get a better understanding of how the Live Book works. Like you, I am either missing something or it looks like the book project needs additional work.

Paul,
If you are using the sim book for diversification purposes, then the fact that you are only picking up the respective equity curves of the underlying portfolios and not the underlying positions is not relevant, since the equity curves are what we are interested in.

In your AAPL example, the portfolio builder may have chosen a high position concentration and low-turnover strategy to minimize taxes or has unwittingly built a flawed strategy that shows extreme concentration. In either case, I think that is an issue that has to be dealt with at the portfolio level, not the book level. In other words, if I put 2 poorly thought out, poorly constructed, highly correlated portfolios into a book and get decent backtest results , then I can probably expect 2 things: (1) the backtest results are due to luck, curve fitting or both, and (2) my results going forward will probably look like nothing in the backtests. However, this is not due to some flaw in the sim book, but rather is a flaw in the strategy/portfolio construction process.

If you have done a decent job in the strategy/portfolio building stage and are looking at things like asset class diversification (e.g., equity ports, fixed income ports, commodity ports, fx ports) or strategy diversification (trend following ports vs mean reversion ports), then, as I stated before, we are only interested in the portfolio equity curves in the sim book, in order to generate a sim book equity curve that we can than run our risk stats on and see if the book makes sense.

Here is where I am having an issue with the live book. Since the sim book is backwards looking, we have the trades and equity curve from the underlying ports. Since the live book is forward looking, we don’t have an equity curve since the trades from the underlying ports haven’t been put on. We won’t have an equity curve until after we put on the trades and see how they work. And since we are running 2 or more ports, there is the issue of reconstitution (making sure the book is following all the trades of the underlying ports) and rebalancing (making sure the weights of the respective portfolios are reset back to the initial settings periodically).

From most of the reading that I have done, it appears that the rebalancing is typically done at quarterly, semi-annual or annual intervals. To make my life simple, I set it annually. On the other hand, the ports typically reconstitute weekly or monthly. If you start running multiple asset ports and start overlaying multiple strategies over that, it starts getting tricky to make sure you are putting on all the trades you should be in order to follow the blended strategy in real time.

I think part of the problem with the live book is that it doesn’t explicitly separate the reconstitution and rebalance functions. Another issue, as Chris points out, is the switch from percentages to dollars when we move from the sim book to the live book. Does it make sense to take a look at what would be involved in taking the sim book and converting that format into a live book and explicitly breaking out the reconstitution and rebalancing functionality?

PS – Sorry for the long post ; it’s Friday and I’m tired.

12

I stopped using live books for this reason, plus the fact that errors (a missed trade or partial fill , partial cancel/resubmit) were much harder to correct.

I “rebalance” now by just withdrawing cash from one port and depositing to another. Since they all have fairly high turnover, they pickup these changes in cash quickly, so it works out OK and it’s simpler.