Diversification, Correlation, and the Limits of Trend FollowingDow Jones Industrial Average IndexTVC:DJIMrRenevDiversification sounds simple in theory: trade enough different things, spread the risk, and losses in one area should eventually be offset by gains elsewhere. In practice, it is much messier, especially for a trend-following strategy. Trend following usually does not rely on a high win rate. Many viable systems seem to operate somewhere around 20% to 40% wins, with profitability coming from asymmetric payouts when a real trend develops. If a strategy wins only one trade out of three but the winners are several times larger than the losers, the system can still have a positive expectancy. The problem is that low win rates naturally produce long losing streaks. At a 33% win rate, fifteen losses in a row are not some impossible statistical freak. Over a large enough number of trades, streaks like that eventually appear. Below roughly 20%, the situation becomes even more difficult because losing sequences can become enormous. A system might still work mathematically, but executing it consistently becomes psychologically and operationally much harder. And this assumes that trades are independent. In financial markets, they usually are not. The problem with correlation If I take five European stock trades within the same few months, those trades may technically involve five different companies, but they are not necessarily five independent bets. Stocks in the same sector are correlated, different sectors are correlated, and entire national stock markets can become correlated when a strong macro force takes over. There is also correlation through time. Market conditions on day one may still be influencing trades taken thirty or fifty days later. If the environment is hostile to a particular strategy, a whole sequence of apparently separate setups can fail for essentially the same reason. This means that a theoretical fifteen-loss streak can turn into something much worse in practice. If several trades are exposed to the same underlying market regime, the losses arrive in clusters. That is why I am skeptical when asset managers talk about holding dozens of genuinely uncorrelated equity positions. Twenty truly independent positions at the same time sounds unrealistic to me. Three to five meaningful sources of independent risk already seems much more plausible. You can diversify equities across industries, countries and market capitalizations, but when there is a serious global repricing, the market has a habit of reminding everyone that they are still equities. Until we can travel through wormholes and trade assets on economically isolated alien planets, perfect diversification remains largely theoretical. Three different kinds of diversification I separate diversification into three broad categories. The first is **instrument diversification**. This includes diversifying across asset classes, geographical markets, sectors and market capitalizations. Stocks, currencies, commodities, bonds, indices and spreads all react to different forces, at least to some degree. The second is **strategy diversification**. This means combining approaches such as trend following, mean reversion, different holding periods or different entry methods. I am personally much less interested in this form of diversification. It sounds attractive on paper, but the more I look at it, the less convinced I am that running several similar trend-following systems adds much value. If they all try to capture roughly the same persistence in prices, their results will often be correlated anyway. Why run several mediocre versions of the same idea if one strategy appears to be better? If I eventually diversify by strategy, I would probably want the large majority of the allocation — perhaps 80% or more — in the system I trust most, leaving a smaller amount for variations, secondary systems or perhaps some mean reversion. I might also use those secondary strategies only when the main system produces very few opportunities rather than forcing myself to trade them constantly. The third form is **operational diversification**. Even a perfectly diversified trading strategy still has another problem if all the capital sits with one broker, one bank or one counterparty. Futures, options and spot markets also involve different operational structures. Spreading capital between institutions does not improve trading expectancy, but it reduces the risk of one external failure becoming catastrophic. That is diversification for survival rather than diversification for returns. My preferred approach: diversify the instruments, not the core idea Instead of constantly adding new strategies, I prefer applying one strategy to as many suitable instruments as possible. A large universe has an important advantage: it allows me to become more selective. If I monitor only fifty stocks, I may feel pressure to take mediocre setups because opportunities are scarce. If I monitor hundreds of stocks, currencies and commodities, I can demand much cleaner conditions and simply reject marginal trades. I can also maintain a secondary watchlist. When the main list is producing plenty of setups, I ignore it. When opportunities become scarce, I can carefully inspect the extended universe rather than weakening my entry criteria. That said, a huge list does not automatically mean a huge amount of diversification. Adding another five hundred stocks does not create five hundred independent opportunities. During a broad crisis, American, European and Japanese stocks can all become exposed to the same global risk-off environment. At that point, adding more equities may do very little. The solution is not necessarily a larger stock universe but a broader set of asset classes. Why wars and crisis periods are difficult Wars are a useful example because they show both the strengths and weaknesses of trend following, and the limits of diversification. At the beginning of a major conflict, many/most markets often become violent and disorderly. Prices gap, volatility rises, liquidity deteriorates and correlations increase. A market can break out sharply, reverse, move again and reverse once more as investors react to headlines, government responses and rapidly changing expectations. No matter how many various stocks you trade, they might all just give false entry signals and all fail because of the uncertainty. This kind of environment can manufacture patterns everywhere. Someone trading double bottoms may suddenly see double bottoms on every chart. Someone trading ABC structures may see ABC structures everywhere. That does not necessarily mean those patterns have become more predictive; increased volatility can simply generate more shapes that resemble them. For trend following, the first phase of a crisis can therefore be extremely awkward. A market makes a huge initial move, the system enters after the breakout, policymakers or investors respond, and price immediately snaps back. Tight stops make the problem worse. But that is only the first phase. Wars can later create some of the strongest macro trends because they alter real economic conditions for months or years. Energy supplies are disrupted, shipping routes change, sanctions redirect trade flows, inflation expectations shift and central banks reprice interest rates. Safe-haven currencies can move strongly, while grains, metals and energy markets may face genuine physical shortages. A common sequence is therefore: **initial shock and whipsaw → repricing → sustained trend** The Russia–Ukraine shock in 2022 is a good modern example. The initial reaction was violent, but the lasting opportunity for diversified trend followers came from persistent moves in commodities, currencies, bonds and other macro markets. For someone trading over several weeks or months, day one may be dangerous while the following months become extremely attractive. The important variable is not “war” itself. What matters is whether the event creates a persistent new regime. Why commodities and currencies are so useful This is one reason I particularly like commodities and foreign exchange. Commodities can have drivers that are almost completely unrelated to the individual-company factors affecting stocks. Oil can trend because of sanctions or OPEC policy. Natural gas can trend because of storage levels or weather. Wheat can trend because of drought. Copper can trend because of Chinese industrial demand. Gold can trend because of monetary expectations or a flight toward perceived safety. These markets are not perfectly independent, but they contain much more genuinely different information than simply adding another national stock exchange. Foreign exchange can also remain highly active when stock markets become difficult. A global crisis may destroy follow-through in equities while simultaneously creating a persistent dollar or yen trend. Investors do not necessarily leave American stocks and buy Australian stocks. Sometimes they leave risky assets altogether, reduce leverage or move toward safer currencies and government debt. That can create a much cleaner trend outside the equity universe. Government bonds and interest-rate markets are interesting for the same reason. In traditional portfolio theory, bonds are valuable when they rise as equities fall. But for a trend follower, that relationship is not even necessary. During an inflationary shock, stocks and bonds may both decline. For a passive investor, that is poor diversification. For a trend follower, a sustained decline in bond prices can still be an excellent opportunity. The direction is secondary. Persistence is what matters. Relative-value spreads may go one step further Spreads and ratios are another area I find potentially interesting. Suppose Brent crude and WTI crude are both choppy, but Brent consistently outperforms WTI. Neither outright chart may provide a particularly clean signal, yet the Brent/WTI (or Brent - WTI) relationship can develop a persistent trend. The same idea can apply to equities. Microsoft and Apple might both decline, but if Microsoft consistently falls less than Apple, the MSFT/AAPL ratio rises. The relative trade is no longer primarily about whether U.S. technology stocks are going up or down. It is about which company is performing better. That potentially removes some of the common market exposure. This may be particularly interesting during crisis periods. Absolute stock trends can become unreliable because the whole market is being pushed around by the same macro uncertainty. Relative trends may survive because one sector, country or company can continue gaining ground against another even while both move erratically in absolute terms. I do not yet know how large these edges are. They may be smaller than outright trend-following opportunities, and spreads introduce their own complications in sizing, volatility and execution. Still, the idea is worth testing because they may provide something normal geographic diversification cannot: a way of extracting trends from relative performance when absolute markets are hostile. The limits of diversification No matter how many instruments I add, I do not think diversification can completely solve the problem of hostile market regimes. There will be periods where correlations rise, volatility becomes violent and every portfolio manager on the planet is constantly repositioning. In those moments, even apparently independent markets may stop giving clean follow-through. There is a hard limit here. More instruments help. More asset classes help. Spreads may help. But sometimes the environment itself simply does not reward trend following. That is why risk management has to assume diversification will occasionally fail. A hypothetical framework could involve risking around 0.5% per trade, limiting exposure to perhaps 1% per correlated cluster, keeping individual asset-class risk below something like 2%, and capping total open risk around 5%. The precise numbers are not important here; the principle is. I also do not want to repeatedly take the same underlying idea. If I already have several positions driven by essentially the same macro factor, another visually different chart may not really be another trade. The same logic applies through time. If a particular market idea has already produced several similar trades within a few months, I should at least be aware that I may be repeatedly betting on the same environment. Drawdown rules can provide another layer of defense. A meaningful drawdown should trigger a review of trade history, correlations and market conditions. A deeper drawdown should lead to reduced risk. At some predetermined point, trading should stop entirely until the cause is understood. The exact thresholds matter less than defining them before the drawdown occurs. No discretionary rebalancing One thing I do not want to introduce is constant discretionary rebalancing. I do not have a team of traders working for me, and I do not want to wake up one morning and suddenly decide that I have “too much” exposure to one area because the recent positions make me uncomfortable. If a setup satisfies the rules, I want to take it. There was no discretionary rebalancing in the backtest, so introducing it live would create a different strategy. If the system produces an uncomfortable cluster of signals, that discomfort is part of the strategy unless I have a predefined risk rule saying otherwise. Skipping trades selectively can be especially dangerous because the trade I decide not to take may be the one that pays for the previous losses. If concentration is genuinely a problem, I would rather solve it systematically through position sizing, cluster limits or instrument selection than by improvising after seeing the signals. Conclusion Diversification is useful, but it does not abolish market reality. A portfolio can contain hundreds of stocks and still behave like one trade during a crisis. Several trend-following systems can still lose together because they are responding to the same underlying price behavior. International diversification can disappear exactly when it is needed most. The goal is therefore not infinite diversification. It's a useful tool, but as usual it's the latest "holy grail" that gets overly hyped way beyond what it actually is. At some point you hit a limit. The goal is to create several genuinely different sources of opportunity while keeping the strategy simple enough to understand and execute. For me, that means concentrating primarily on one trend-following approach and applying it across a broad universe of stocks, currencies, commodities, bonds, indices and potentially relative-value spreads. A small amount of strategy diversification may eventually make sense, but I see much more value in expanding the opportunity set than in constantly inventing new systems. There will still be periods when markets become violent, correlated and directionless, and nothing follows through properly. No portfolio construction trick can guarantee protection from that. The objective is not to eliminate those periods. Something that is "all weather" either catches 0 alpha, or very little, or is a lie, in all cases it's a marketting ploy. The key is to survive crisis without damaging the strategy so badly that I am no longer there when the next persistent trend appears.