Aggressive rebalancing

[rukawa]

I was playing around with net-net screens with EQS on Bloomberg. The default on Bloomberg backtest function is monthly rebalance which I didn't realize. I ran the screen over the period from 2000-2017 and got some insane result: 70000% increase. An average of 50% returns over the period. I thought shit this result must be garbage resulting from data errors so I filtered out stocks with less than 1 million mcap and less than 1 cent prices. Still I was getting the same number. Then finally I figured out that the rebalance was monthly and switched to annual.

 

The result was a more reasonable 35% a year which exactly agrees with the net-net literature (Oppenheimer, Tobias Carlisle). I knew this was a reasonable result. But here is the thing I don't think the monthly rebalance result was wrong. I think its real. Its been found with equally weighted indexes too that re-balancing monthly is what drives returns. Its also interesting to observe that the graph of returns I observed was very smooth...much more smooth than yearly re-balancing. There were less draw-downs, less flat periods...it just looked like a nice smooth exponential curve.

 

Its also interesting to think about what is going on here. Because net-nets don't really change in value that often. There are probably only a small percentage of stocks that will increase substantially in any 1 year but it appears that selling these promptly and re-balancing is a fairly effective way of boosting returns.

 

The real question is whether this higher return can be captured. I believe it can be without too much additional cost and that the effect is significant.

[frommi]

Does the bloomberg backtest use realistic transaction costs and slippage? With netnets slippage can be the deciding factor if more or less rebalancing is better. Also the tax consequences of monthly rebalancing can ruin your backtest. 1% per month easily eats the excess return of a monthly rebalance. (And maybe you can do this with a tiny amount of money, but i don`t think you will be able to monthly rebalance a netnet portfolio with more than 500k in it.)

 

The best i was able to get in a backtest with portfolio123 was a CAGR of 51%, but the drawdown was 68%. It was an equal weight simulation of the 10 netnets with the highest discount to tangible book. With 20 stocks it looked more "useable", since it was still a CAGR of 44% with market like drawdowns. I used 0.5% as commission and 0.5% as slippage which is still very optimistic for a netnet-system in my view. That system used a rule to sell at NCAV, so no fixed holding period. (either NCAV deterioated very fast or the stocks go up to NCAV, i don`t think there was a stock longer than 3 years in the portfolio.)

[Guest]

is there a way to access the list without a bloomberg or capital iq subscription?

[rukawa]

Does the bloomberg backtest use realistic transaction costs and slippage?

 

No. In Bloomberg AFAIK, both these would be zero. The buy/sell criterion is buy when mcap is less than 2/3rds NCAV and sell when above this. Basically you can think of bloomberg as liquidating the portfolio each month and buying each stock satisfying the screen (mcap < 2/3rds NCAV) each month. However there is no reason for anyone to do this in practice. In practice you could keep the whole portfolio intact and only buy/sell those stocks that have moved significantly.

 

Also the tax consequences of monthly rebalancing can ruin your backtest. 1% per month easily eats the excess return of a monthly rebalance....I used 0.5% as commission and 0.5% as slippage which is still very optimistic for a netnet-system in my view.

 

I don't know that it is optimistic. In fact it seems really high to me to assume 1% total costs monthly. I don't see why slippage would be 0.5% of the whole portfolio in one month. You aren't trading the whole portfolio each month in practice.  I would guess you are trading less than 5% of all stocks. The only way to get 0.5% on a portfolio basis for one month would then be to have slippage of 10% on a position basis which seems really high. I also don't get how its possible to get 0.5% commissions, even on a per position basis. I can and have bought net nets in Japan, Singapore, USA and Canadian Venture exchange. Generally you are paying around 0.25%. On questrade I recently bought $3000 of a US stock and paid $10. Which is $10/$3000 = 0.33%. In Japan I have looked at all my trades in interactive brokers over 2017 and I see that I bought $31478 and paid $25.23 in commissions across 9 trades which means I paid 0.08%.

 

The worst, worst, is the Canadian Venture Exchange where I pay around 0.5% and sometimes even more (Russell Breweries!!). But this is all on a per position basis. On a portfolio basis if you are trading 5% of the portfolio each month and you paying 0.5% on a position, your cost across the whole portfolio is 0.025% or 2.5 bps per month. In 1 year that means 2.5*12= 30bps or 0.3%. I just don't get how that can eat up the additional 15% or 1500 bps you are getting from monthly rebalance.

 

For tax consequences, a 50% return a year assuming a 25% capital gain tax and 100% annual turnover, gets bumped down to 37.5%. With 50% annual turnover, your after tax return is 50%*(1-turnover%*cap gain tax rate) = 50%*(1-0.5*0.25)=43.75% and with 33% turnover your return is 45.6%. So even at 100% turnover, just taking taxes alone you are still better off with monthly rebalance but at lower rates of turnover you are far better off. I don't think 50% turnover is unreasonable and at this level you are still getting nearly 9% of extra return.

 

Of course I am assuming a certain story for what is going on and my story almost certainly wrong. It could be that you are actually trading much more than 5% of the portfolio in each month and that what is really happening is that large percentage of marginal stocks are bouncing around the buy/sell criterion for the screen and coming in and out of the portfolio repeatedly. My assumption is more along the lines of there are a few stocks which have big sudden runups. I'll have to investigate further to figure out what is driving this difference and how exploitable it is.

 

he best i was able to get in a backtest with portfolio123 was a CAGR of 51%, but the drawdown was 68%. It was an equal weight simulation of the 10 netnets with the highest discount to tangible book.

 

That's pretty impressive. I tried all kinds of things and found it very difficult to get much higher than 35% without monthly rebalance. Considering your test included transaction costs and slippage I am surprised you could get that.

[rukawa]

is there a way to access the list without a bloomberg or capital iq subscription?

 

Sorry not that I know of :(.

 

Bloomberg actually tells you that that the datasets are point in time without survivorship bias which shocked me. Those datasets are tremendously expensive and only CRSP/Compustat has these AFAIK. Maybe Bloomberg is getting them from there.

[frommi]

I looked at my account statements and you are right, trading japanese stocks was pretty cheap, but trading otc stocks with very low prices and buying australian stocks was expensive for me. (around 0.5% per trade, so i took that as a reference.) And when you don`t cross the ask its sometimes hard to build meaningful positions, thats the reason i think that 0.5% slippage in a backtest is reasonable. I missed the point of not rebalancing the whole portfolio so you are probably right that it doesn`t lower the returns so much.

 

But a part of whats going on with monthly rebalancing might be that you are capturing the bid-ask spreads when the closing prices bounce between them.

 

@stahlehyp you can start a portfolio123.com trial to get lists of netnets.

[frommi]

Here are the results of my tests:

 

10 stocks:

https://drive.google.com/open?id=0BzQbS-AUNeo9OUt2NEZHcHpaM1U

 

20 stocks:

https://drive.google.com/open?id=0BzQbS-AUNeo9ZDRPaGJQeXJISWs

 

buy rules:

ncav/mktcap > 1.4

price > 0.1

avg. volume10day > 0.1

mktcap < 150 million

stable sharecount

ncav burnrate < 25% YoY and QoQ

no biotech,financials,o&g stocks.

 

ranked by tangible book

 

sell rule:

ncav < mktcap

[NormR]

I'm not saying that Net-Nets are bad.  But ...

 

Backtesting microcaps can be problematic due to liquidity.  Bid/ask bounce can skew the test.  Getting in/out in volume is an issue.  There can be big data quality issues in the micro space.  etc.

 

I've seen huge spreads (i.e. bid: $8 ask:$10) on small stocks on the TSX in the past with a few hundred shares traded a day - on a heavy trading day.   

 

There will also be issues of getting a big basket of Canadian net-nets in many non-bear market years.

 

Frankly, I'm skeptical of backtests with CAGRs north of ~20%/yr for long periods.  It usually indicates some problem with the test.  i.e. data mining, etc.

 

Healthy skepticism along with a good deal of sensitivity testing is advised.  If it's too good to be true...   

 

Add: Btw, go to the "counts" tab on the Bloomberg report to look at the backtest's turnover.

[Guest]

rukawa and frommi, thanks for the help!

 

 

have you guys tried any of these strategies in real life?

[Patmo]

I've only ever invested in what everybody here considers crappy businesses and whatnot, netnets, unprofitable, liquidations, distressed m&a and similar types of things. In my experience you will do way better by not getting married to an idea. Sell it as soon as some catalyst plays out, or share price pops a bit unexpectedly. You'll get many, many winners that would've otherwise been stagnant (or losers), and do way better than hanging onto everything for that one idea that might become a monster bagger. With those investments, you have to play small ball for best results. At least it's my experience. So those backtest results don't surprise me so much...

[rukawa]

rukawa and frommi, thanks for the help!

have you guys tried any of these strategies in real life?

 

I'm not saying that Net-Nets are bad.  But ...

 

Backtesting microcaps can be problematic due to liquidity.  Bid/ask bounce can skew the test.  Getting in/out in volume is an issue.  There can be big data quality issues in the micro space.  etc.

 

I've seen huge spreads (i.e. bid: $8 ask:$10) on small stocks on the TSX in the past with a few hundred shares traded a day - on a heavy trading day.   

 

There will also be issues of getting a big basket of Canadian net-nets in many non-bear market years.

 

Frankly, I'm skeptical of backtests with CAGRs north of ~20%/yr for long periods.  It usually indicates some problem with the test.  i.e. data mining, etc.

 

Healthy skepticism along with a good deal of sensitivity testing is advised.  If it's too good to be true...   

 

Add: Btw, go to the "counts" tab on the Bloomberg report to look at the backtest's turnover.

I have just started this strategy about 8 months ago. Right now its too soon to tell.

 

The only guy I know that tried this in real life and reported performance is Ben Graham. He was reporting about 20% a year. However he filtered for positive earnings and dividends, which according to some studies reduces returns. Tobias Carlisle reports 1.96% per month for positive prior year earnings and 3.38% per month for negative and 2.55% across the whole sample. This translates to 26% per year and 50% per year and 35% per year respectively. Though when I tried to filter like this I didn't see any difference between pos and neg earning net-nets.

[rukawa]

Frankly, I'm skeptical of backtests with CAGRs north of ~20%/yr for long periods.  It usually indicates some problem with the test.  i.e. data mining, etc.

 

Healthy skepticism along with a good deal of sensitivity testing is advised.  If it's too good to be true... 

 

The 50% does appear to be too good to be true. My guess would be that at minimum net-nets provide 20% a year in a real life implementation based on Ben Grahams reports and the fact that nearly ever study has reported at least this performance. The 30%-50% range is more iffy. But if its wrong I would like to know exactly what is wrong. All the issues you pointed out like insufficient volume, low liquidity are things we can test for since volume is a measure in Bloomberg. I can filter for something like volume*price > $10000 cad daily.

 

Bid/ask bounce can skew the test.  Getting in/out in volume is an issue.  There can be big data quality issues in the micro space.  etc.

 

The bid/ask is what everyone states as an objection but its one that doesn't make sense to me. Any backtest is based on close price. So this already includes the bid ask/spread...a close price implies a bid and an ask at that price at that time for a trade. To be honest I have found with anything over about 4 million mcap I can almost always get my trade filled at the last close. Even under 4 million....lets say 1-2 mil I still get last close, it just takes a longer. Its only sub 1 million or dark companies which are no longer reporting where in practice I see real issues.

 

There will also be issues of getting a big basket of Canadian net-nets in many non-bear market years.

 

Yes the same objection is given in US...that there aren't enough US net nets. My solution is to search internationally...Japan, Australia, Europe etc. Generally there are enough net-nets internationally to compile a good list.

[Guest]

Carlisle runs a fund. It looks like it was shut down/name change into something else (was Eyquem and now is Carbon Beach). According to Linkedin he started Carbon Beach in 2010. However, if  you look at the shareholder letters they just have returns from 2016-2017 (the first year was up a little over 50%). It's quite strange that he'd change the name and or get rid of the previous track record if his performance was any good.

[frommi]

rukawa and frommi, thanks for the help!

have you guys tried any of these strategies in real life?

 

Yes, i am running 50% of my money that way but only since 12 months.

[Pelagic]

Frommi in the Google Drive links you posted it shows that the rebalance frequency is weekly, is that correct and did you notice weekly performed better over the same period than monthly/yearly?

[frommi]

Frommi in the Google Drive links you posted it shows that the rebalance frequency is weekly, is that correct and did you notice weekly performed better over the same period than monthly/yearly?

 

Since it is a portfolio simulation, that is the frequency the engine uses to evaluate if it has to execute a buy or sell-rule. Stocks are only bought or sold if the rules are met. Yearly or monthly rebalancing is worse, because the simulation can`t take advantage of price drops/surges in between that timespan. I checked the transaction log after the simulation and it was the way i expected this to work.

[NomadicRiley]

Here are the results of my tests:

 

10 stocks:

https://drive.google.com/open?id=0BzQbS-AUNeo9OUt2NEZHcHpaM1U

 

20 stocks:

https://drive.google.com/open?id=0BzQbS-AUNeo9ZDRPaGJQeXJISWs

 

buy rules:

ncav/mktcap > 1.4

price > 0.1

avg. volume10day > 0.1

mktcap < 150 million

stable sharecount

ncav burnrate < 25% YoY and QoQ

no biotech,financials,o&g stocks.

 

ranked by tangible book

 

sell rule:

ncav < mktcap

 

Frommi,

 

Curious as to how you were able to get such promising results.  I also have a portfolio123 and just tried a similar simulation using your rules and the results were abysmal (negative annual returns)

 

This is my attempt to mimic your rules above:

 

Buy Rules:

($NCAV2/MktCap) > 1.4

Price > 0.1

Vol10DAvg > 0.1

MktCap < 150

(SharesFDQ / SharesFDPY) < 1.2

($NCAV2/$NCAV2PQ) < 1.25

($NCAV2/$NCAV2PQY) < 1.25

 

Sell Rules:

$NCAV2 < MktCap

 

Definitions:

$NCAV2 = AstCurQ - LiabCurQ - PfdEquityQ

 

Universe:

All Stocks USA

Exclude GICS (352010 - Biotech, 40 - Financials, 10102020 - O&G Exploration)

 

 

[TwoCitiesCapital]

I put a small portion of my money into a "passive" P/B portfolio. While I agreed with all the research surrounding the topic, I started thinking that an annual re-balance may not make sense. Looking at the data, low P/B, P/S, and P/E portfolio's tended to outperform over a 12 month horizon, but if held static, tended to under perform over 2- and 3- year horizons (and longer).

 

This makes intuitive sense in that you're getting the "last puff" out of cigar butt. So, in my hypothesis, I figured you have to constantly turn cigar butts to get the equivalent of a cigar. A portfolio that turned modestly more frequently should have a good chance of outperforming if you're getting rid of the new "dead weight" from recent out performers and turning those into more potential puffs more frequently. 

 

I started turning my portfolio anytime a stock hard a large run-up that put it within 10-20% of its book value. If it ran up - I sold regardless of the time frame held (tax deferred account so tax implications do not matter).

 

Some stocks were sold tremendously early and I missed out on a good run. Other stocks were sold after a 50-60% bounce which proved just that - a bounce - before returning to the the prices I had paid and lower.

 

I wasn't immediately re-investing the proceeds, as I did not have the time to run the screen, nor have I accounted for what the actual results would have been had I held, but based on what I know of the names bought/sold and their performance, it seems that I likely that I have done better than a buy/hold strategy in the same names over the past 2-years.

[JayGatsby]

Can anyone share a list (or parts of a list) of what this portfolio would currently hold? I ran a similar screen on CapitalIQ (https://drive.google.com/open?id=0B1tb4Z-iuO9Bd3NDYWxmT3JOYWc). I've looked through a good number of the results and haven't found any that really jump out at me as good investments. That has the bias of my opinion, so the key to making the strategy work may be to eliminate that  ;). Some are Chinese companies, some are retailers with deteriorating results, some are homebuilders/land owners, some don't have English financials (which leaves you betting on the CapIQ/Bloomberg analyst inputting the numbers correctly).  Trans World Entertainment is one I've watched for a while... they have a large shareholder, a lot of inventory and a huge amount of NOLs. If they just slowly liquidated stores as sales declined and reinvested that into something low risk but cash producing it would be a great investment. Instead they overpaid for an Amazon seller that loses money.

 

My personal experience has been that investing primarily on the basis of low price/book / net-net has generated my worst results. Again, that's a limited sample set which incorporates my selection bias. I'm skeptical of a company that can't earn decent ROIC in today's economy, unless it's O/G (or something else that is cyclically down).

[TwoCitiesCapital]

Can anyone share a list (or parts of a list) of what this portfolio would currently hold? I ran a similar screen on CapitalIQ (https://drive.google.com/open?id=0B1tb4Z-iuO9Bd3NDYWxmT3JOYWc). I've looked through a good number of the results and haven't found any that really jump out at me as good investments. That has the bias of my opinion, so the key to making the strategy work may be to eliminate that  ;). Some are Chinese companies, some are retailers with deteriorating results, some are homebuilders/land owners, some don't have English financials (which leaves you betting on the CapIQ/Bloomberg analyst inputting the numbers correctly).  Trans World Entertainment is one I've watched for a while... they have a large shareholder, a lot of inventory and a huge amount of NOLs. If they just slowly liquidated stores as sales declined and reinvested that into something low risk but cash producing it would be a great investment. Instead they overpaid for an Amazon seller that loses money.

 

My personal experience has been that investing primarily on the basis of low price/book / net-net has generated my worst results. Again, that's a limited sample set which incorporates my selection bias. I'm skeptical of a company that can't earn decent ROIC in today's economy, unless it's O/G (or something else that is cyclically down).

 

The studies have actually shown that it IS the worst of the worst that provide the best returns. Narrowing the focus of low P/B stocks by selecting favorable metrics like positive cash flows, positive earnings, and etc has only ever impaired the long-term results of the strategy. You actually get better returns by focusing only on companies that are LOSING money. You truly have to buy the nonsensical/hated over-levered equity stubs where a handful will surprise and return 50-250% while everything else languishes for 12 months.

 

The drawdowns for this type of strategy are as high as 90% in a given year. That is why few people have a stomach for this - you can't justify why you hold them and the volatility/drawdowns are insane when a bad year hits. This is exactly why the opportunity exists as well. As simple as it is to implement, it's not easy to do so.

[JayGatsby]

The studies have actually shown that it IS the worst of the worst that provide the best returns. Narrowing the focus of low P/B stocks by selecting favorable metrics like positive cash flows, positive earnings, and etc has only ever impaired the long-term results of the strategy. You actually get better returns by focusing only on companies that are LOSING money. You truly have to buy the nonsensical/hated over-levered equity stubs where a handful will surprise and return 50-250% while everything else languishes for 12 months.

 

The drawdowns for this type of strategy are as high as 90% in a given year. That is why few people have a stomach for this - you can't justify why you hold them and the volatility/drawdowns are insane when a bad year hits. This is exactly why the opportunity exists as well. As simple as it is to implement, it's not easy to do so.

That's really interesting. So the key is the companies that are so beat up that no rational person would touch them... with high TBV the upside optionality is higher than the downside optionality (on average).

[frommi]

Frommi,

 

Curious as to how you were able to get such promising results.  I also have a portfolio123 and just tried a similar simulation using your rules and the results were abysmal (negative annual returns)

 

This is my attempt to mimic your rules above:

 

Buy Rules:

($NCAV2/MktCap) > 1.4

Price > 0.1

Vol10DAvg > 0.1

MktCap < 150

(SharesFDQ / SharesFDPY) < 1.2

($NCAV2/$NCAV2PQ) < 1.25

($NCAV2/$NCAV2PQY) < 1.25

 

Sell Rules:

$NCAV2 < MktCap

 

Definitions:

$NCAV2 = AstCurQ - LiabCurQ - PfdEquityQ

 

Universe:

All Stocks USA

Exclude GICS (352010 - Biotech, 40 - Financials, 10102020 - O&G Exploration)

 

Exact rules:

https://drive.google.com/open?id=0BzQbS-AUNeo9UHpjZm1aaks3a00

 

exclude china and sharecount 5% allowed difference instead of 20% may be the drivers.

[NomadicRiley]

 

Exact rules:

https://drive.google.com/open?id=0BzQbS-AUNeo9UHpjZm1aaks3a00

 

exclude china and sharecount 5% allowed difference instead of 20% may be the drivers.

 

Thanks for the screenshot, I was able to get very similar #'s using those rules!

[NomadicRiley]

Can anyone share a list (or parts of a list) of what this portfolio would currently hold? I ran a similar screen on CapitalIQ (https://drive.google.com/open?id=0B1tb4Z-iuO9Bd3NDYWxmT3JOYWc). I've looked through a good number of the results and haven't found any that really jump out at me as good investments. That has the bias of my opinion, so the key to making the strategy work may be to eliminate that  ;). Some are Chinese companies, some are retailers with deteriorating results, some are homebuilders/land owners, some don't have English financials (which leaves you betting on the CapIQ/Bloomberg analyst inputting the numbers correctly).  Trans World Entertainment is one I've watched for a while... they have a large shareholder, a lot of inventory and a huge amount of NOLs. If they just slowly liquidated stores as sales declined and reinvested that into something low risk but cash producing it would be a great investment. Instead they overpaid for an Amazon seller that loses money.

 

My personal experience has been that investing primarily on the basis of low price/book / net-net has generated my worst results. Again, that's a limited sample set which incorporates my selection bias. I'm skeptical of a company that can't earn decent ROIC in today's economy, unless it's O/G (or something else that is cyclically down).

 

I ran the screen frommi described above against the Portfolio123 data today and it is only returning 2 current symbols.

 

GIGM GigaMedia LTD MktCap ~33m and NCAV ~57m

MSN Emerson Radio Corp MktCap ~35m and NCAV ~52m

 

 

[frommi]

Can anyone share a list (or parts of a list) of what this portfolio would currently hold? I ran a similar screen on CapitalIQ (https://drive.google.com/open?id=0B1tb4Z-iuO9Bd3NDYWxmT3JOYWc). I've looked through a good number of the results and haven't found any that really jump out at me as good investments. That has the bias of my opinion, so the key to making the strategy work may be to eliminate that  ;). Some are Chinese companies, some are retailers with deteriorating results, some are homebuilders/land owners, some don't have English financials (which leaves you betting on the CapIQ/Bloomberg analyst inputting the numbers correctly).  Trans World Entertainment is one I've watched for a while... they have a large shareholder, a lot of inventory and a huge amount of NOLs. If they just slowly liquidated stores as sales declined and reinvested that into something low risk but cash producing it would be a great investment. Instead they overpaid for an Amazon seller that loses money.

 

My personal experience has been that investing primarily on the basis of low price/book / net-net has generated my worst results. Again, that's a limited sample set which incorporates my selection bias. I'm skeptical of a company that can't earn decent ROIC in today's economy, unless it's O/G (or something else that is cyclically down).

 

I ran the screen frommi described above against the Portfolio123 data today and it is only returning 2 current symbols.

 

GIGM GigaMedia LTD MktCap ~33m and NCAV ~57m

MSN Emerson Radio Corp MktCap ~35m and NCAV ~52m

 

Yes, there is no way around international markets if you want to build a diversified netnet portfolio at the moment. The most stocks in my portfolio are currently from Japan, Singapur, Hongkong and Poland.