The two previous articles described a breakout book and a cross-sectional momentum book on Borsa İstanbul. They have almost the same record. One returned 139.9% a year with a 43.4% drawdown, the other 115.4% with 42.1%.
Given a choice between two strategies with nearly identical returns and nearly identical drawdowns, the interesting move is not to choose. This article is what happens when both run at once, and specifically what the combination buys that neither has alone.
The construction
Equal capital. Each sleeve gets 1,000,000 lira, they run independently with no shared cash pool, and the combined equity is the sum of the two. Nothing is rebalanced between them, so their weights drift as they perform differently.
Costs are identical at 0.5% per fill in both sleeves, which matters more than it sounds: an uncosted sleeve appears to add free return to a blend, and any contribution analysis built on that is meaningless.
The two books share a universe and nothing else:
| Train | Holdout | |||||||
|---|---|---|---|---|---|---|---|---|
| CAGR | Max DD | Sharpe | MAR | CAGR | Max DD | Sharpe | MAR | |
| UGUnusual Gain, breakout | 139.9% | -43.4% | 2.57 | 3.22 | – | – | – | – |
| MLMomentum Leaders, ranked | 115.4% | -42.1% | 2.55 | 2.74 | – | – | – | – |
One is event-driven: it acts when a stock makes a new high and turns its positions over in about 22 days. The other is a calendar-driven ranking: it acts on a fixed cycle regardless of events and turns over in about 56 days. They are different mechanisms pointed at the same market.
What the combination does
- CAGR
- 129.5%
- Max DD
- -41.6%
- Sharpe
- 2.86
- MAR
- 3.11
Compare each number to the two sleeves.
| Unusual Gain | Momentum Leaders | Average of the two | Combined | |
|---|---|---|---|---|
| Return | 139.9% | 115.4% | 127.7% | 129.5% |
| Worst drawdown | -43.4% | -42.1% | −42.8% | -41.6% |
| Volatility | 36.2% | 31.6% | 33.9% | 31% |
| Sharpe | 2.57 | 2.55 | 2.86 | |
| Return over drawdown | 3.22 | 2.74 | 3.11 |
Start with the return, because it is the one number that has to behave. The combined book ends at 259.4× its starting capital, and the two sleeves end at 348.9× and 169.8×. Average those and you get 259.4×. That is not a result, it is an arithmetic check: equal capital with no shared pool means the combined book is the average of the two, and a combined multiple that came out anywhere else would mean a mistake somewhere. The annual rate reads slightly above the average of the two rates (129.5% against 127.7%) for the same reason a rate is not a linear function of a multiple, not because anything was gained.
What the combination actually buys shows up in the denominator, and it is smaller than the story usually told about diversification. Volatility falls from an average of 33.9% to 31%, roughly three points. The worst drawdown falls from an average of 42.8% to 41.6%, a gain of 1.2 points, which is real and is not much.
So the risk-adjusted measures split. Sharpe lands above either sleeve, at 2.86 against 2.57 and 2.55, because volatility is what Sharpe is made of and volatility is where the blending worked. Return over drawdown does not: 3.11 sits between the two sleeves rather than above them, below the breakout sleeve's 3.22, because the worst day for this pair was a day both of them were falling.
That distinction is the whole article. Blending two books like these smooths the ordinary weeks. It does not do much about the week you actually care about.
The chart that actually matters
The equity chart above is unremarkable, because averaging two similar curves produces a curve between them. The drawdown chart is where the argument lives.
How correlated are they, really
The daily return correlation between the two sleeves is 0.466.
That number deserves care. It is not low. Two long-only momentum books fishing in the same 678 Turkish stocks are going to move together most of the time, and a correlation near 0.5 says exactly that: about half of what happens to one happens to the other.
Which is why the drawdown benefit came out at 1.2 points rather than something impressive. This is not a diversified portfolio. It is one bet on Turkish equity momentum, expressed two ways, and the second way removes some of the noise around the first without changing what the position is.
The rolling picture is the honest one. Correlation between these two ranges from below 0.2 to above 0.7 depending on the period, and the direction of that variation is unhelpful: it tends to be highest when both books are falling, which is when the diversification is most needed. A single headline correlation of 0.466 hides that entirely, and it explains the 1.2-point drawdown result better than the average does: at the moment the combined book set its worst level, these two were not behaving like separate strategies at all.
Year by year
| Year | J | F | M | A | M | J | J | A | S | O | N | D | Year |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2020 | -19 | -17 | 40 | -1 | 4 | 23 | -8 | 38 | -4 | 19 | 34 | 160% | |
| 2021 | 5 | 9 | 15 | -15 | -10 | -6 | -3 | 2 | -6 | 9 | 53 | -16 | 24% |
| 2022 | 10 | -11 | 21 | 7 | 14 | 12 | 17 | 27 | 6 | 17 | 32 | 10 | 331% |
| 2023 | -17 | -5 | 3 | -2 | 8 | 56 | 19 | 5 | 29 | 12 | 5 | -17 | 109% |
| 2024 | 33 | 2 | 0 | 4 | -6 | 3 | 6 | 13 | -2 | -3 | 3 | -5 | 50% |
| 2025 | -7 | 23 | 19 | 44 | -21 | 15 | 17 | 24 | -10 | 7 | -7 | 6 | 151% |
| 2026 | 6 | 8 | 20 | 17 | 3 | 19 | 12 | 10 | -4 | 128% |
Monthly return in percent. Colour saturates at 40%.
Set the combined years against the sleeves:
| Year | Unusual Gain | Momentum Leaders | Combined | BIST 100 |
|---|---|---|---|---|
| 2021 | 14.84% | 49.6% | 23.98% | 17.58% |
| 2022 | 385.99% | 214.17% | 331.03% | 221.47% |
| 2023 | 125.27% | 54.94% | 108.95% | 25.74% |
| 2024 | 54.24% | 29.6% | 49.81% | 17% |
| 2025 | 109.8% | 370.48% | 151.15% | 13.43% |
| 2026 | 116.85% | 154.27% | 127.82% | 14.64% |
Two rows carry the argument for holding both. In 2023 the breakout book made 125.27% while the ranking book made 54.94%. In 2025 it reversed: the ranking book made 370.48% while the breakout book made 109.8%. The leadership swaps, and a blend collects a decent year in both cases instead of an outstanding year followed by a mediocre one.
2025 is also the row that should temper all of it, and for a reason this article has so far let pass without comment. The combination made 151.15% that year. An evenly split blend of the same two sleeves would have made 240.1%.
Equal capital is set once, at the beginning, and never restored. Whichever sleeve compounds faster accumulates weight, and the breakout sleeve pulled ahead immediately. It made 285.73% in 2020 against 34.56%, so the book was already 74.1% breakout by the end of the first year and never came back. By the end of 2024 it was 84.5%. When the ranking sleeve finally had its year, the book barely owned it.
That is diversification decaying on its own. The whole argument for holding both books rests on neither one dominating, and a never-rebalanced blend spends most of its life drifting away from the split that produced the benefit. That is part of why the drawdown benefit came out at 1.2 points rather than the number two half-correlated sleeves at a true 50/50 would give. Rebalancing back to equal capital would fix it and would also introduce a decision that has not been tested, which is why the figures here describe the book as it actually ran.
What it is measured against
The market column is not the published index. The exchange publishes index membership at quarterly reviews, and that column holds exactly the names in BIST 100 during each review period, equally weighted, rebalanced only on review dates, paying the same 0.5% per fill the sleeves pay.
| CAGR | Max DD | Sharpe | MAR | |
|---|---|---|---|---|
| Combined book | 129.5% | -41.6% | 2.86 | 3.11 |
| BIST 100 basket, equal weight | 48.8% | -36.1% | 1.59 | 1.35 |
| BIST 100 basket, traded-value weight | 47.1% | -38.3% | 1.46 | 1.23 |
| BIST 30 basket, equal weight | 48.2% | -35.3% | 1.49 | 1.37 |
| Whole universe, equal weight | 73.3% | -40.2% | 2.27 | 1.82 |
Equal weight and traded-value weight land within about two points of each other, so the comparison does not turn on which approximation I chose. The row worth arguing with is the last one: an equal-weight basket of all 678 names returned 73.3% a year, ahead of either index, because small caps beat large caps badly here. Both sleeves fish in that whole universe, so some of what looks like skill is the pond.
What I would want before trusting this
- Everything from the two source articles applies. Both books were parameterised on this window, both records rest heavily on one exceptional year, and neither has seen a sustained bear market. Combining two in-sample strategies produces an in-sample combination.
- Equal capital was chosen, not derived. I have not swept the weighting, and I deliberately have not, because tuning the blend on the same window that produced the sleeves is how a third layer of overfitting gets added to two existing ones. Equal capital is a prior, not a result.
- The blend decays toward its winner. Equal capital is never restored, so the split drifts with performance and the diversification benefit drifts with it. 2024 shows what that costs. Any real version of this needs a rebalancing rule, and that rule needs testing on something other than this window.
- The correlation is the fragile part. It averages 0.466 and climbs above 0.7 in the periods that matter most. The drawdown benefit measured here is what happened, not what is guaranteed.
- Capacity compounds. Two small-cap Turkish books running at once compete for the same names on the same days. Neither article has addressed capacity properly, and running them together makes it worse rather than better.
The honest summary is narrower than the one this article set out to write: two strategies with nearly the same return and nearly the same drawdown, run together, gave up nothing in return, took about three points off volatility, and took 1.2 points off the worst drawdown. Sharpe improves on either sleeve; return over drawdown does not. That is worth doing and it is close to the smallest version of worth doing. It is not a portfolio, and at 0.466 correlation between two long-only momentum books in one market, it was never going to be.