On February 19, 2020 the S&P 500 closed at a record 3,386. A simple trend model run on that index, the kind of rule commodity trading advisors use, was 104% long. Fifteen trading days later, on March 11, the same model was 33% short, and nobody had to change their mind: falling prices flipped its signals and soaring volatility shrank its size. That is what strategists mean by “CTA selling”, and it is why traders watch CTA positioning. If CTAs are new to you, start with managed futures explained.
What is CTA positioning?
CTA positioning is the aggregate long or short exposure that trend-following funds hold in futures markets such as the E-mini S&P 500, Treasury futures, currencies and commodities. A commodity trading advisor is a firm registered with the CFTC through the National Futures Association to advise on or manage futures trading; many of the largest follow trends.
The market cares because trend followers are price-insensitive and rule-bound. A discretionary manager might buy a dip because valuations look attractive. A trend model sells it because its look-back returns turned negative, and it does so in size, across many funds running similar rules. That makes CTAs one of the main sources of mechanical, non-fundamental flows, alongside volatility control funds, quarter-end pension rebalancing and options dealer hedging.
Nobody publishes CTA positions daily. So “CTA positioning” in a strategist note is nearly always one of two things: a reading of the CFTC’s weekly data, or the output of a model that mimics trend followers. This guide covers both.
How do trend models turn price into exposure?
Trend models turn price into exposure in two steps: a signal that sets direction, and a volatility scalar that sets size. Multiply the two and you have the position.
The signal comes from comparing today’s price with past prices. Moskowitz, Ooi and Pedersen’s time series momentum study used the past 12-month return across 58 futures markets; real programs blend several look-backs. Ours averages the sign of the 1, 3 and 12-month returns, so it can be +1 (all up), +1/3, −1/3 or −1 (all down).
The size comes from volatility. To keep risk roughly constant, the position is multiplied by target volatility divided by recent realized volatility. With a 10% target and the S&P 500 realizing 10%, the scalar is 1. If realized volatility doubles to 20%, it halves to 0.5. We cap it at 1.5 so the model never runs more than 150% long or short.
Run on FRED’s daily S&P 500 closes, the model behaves the way CTA commentary describes. It was long on about 78% of days from October 2017 to October 2026, often near its cap in calm rallies. Then it flipped within weeks of each major selloff:
- Late 2018. From +1.41 on October 4, 2018 to −0.33 a week later, and about −0.46 by Christmas Eve.
- March 2020. From +1.04 on February 19 to +0.25 on February 26 and −0.33 on March 11.
- 2022. From +0.73 on January 3 to −0.22 by January 25, then about −0.41 in early May as all three look-backs turned negative.
- Spring 2025. From +0.80 on February 19 to −0.23 by March 5.
On October 6, 2026, the last close in our data, it was back at +0.90.
What are CTA trigger levels?
CTA trigger levels are the index prices at which a trend model’s signal would change. For a look-back rule they are simply old prices: the model turns less bullish when today’s close drops below where the index stood one, three or twelve months ago.
On October 6, 2026 the S&P 500 closed at 7,818.93. In our model the first trigger was the close of a month earlier, 7,719, about 1.3% below: a drop under it would cut the signal from +1 to +1/3. The three-month trigger sat at 7,483, about 4.3% below, where the signal would turn negative. The 12-month trigger was 6,716, more than 14% below, where the model would be fully short.
Bank research desks publish the same kind of ladder, usually with more look-backs and with moving averages or breakout channels instead of simple returns. A moving-average trigger fires when the price crosses the average of the last n closes; a breakout trigger fires when the price makes a new n-day high or low. They share three properties:
- They roll. Tomorrow the reference price is a different day’s close, so the level moves even if the market doesn’t.
- The short-term level is usually close. The one-month trigger is often within a few percent of the price, so small dips trim exposure.
- The long-term level does the damage. When the slow signal flips, the whole model is on the same side, which is when headlines about CTAs “flipping short” appear.
Why does CTA selling accelerate in a selloff?
CTA selling accelerates because both multipliers fall at the same time: each new low turns another look-back negative, and the rising volatility that comes with a selloff shrinks position size. One effect alone would cause gradual selling; together they compound.
Here is how the 2020 move breaks down in our model. The dollar column applies the exposure change to a hypothetical $10 billion of trend-following capital.
| Date (2020) | S&P 500 close | Signal | Vol scalar | Exposure | Change on $10B (illustrative) |
|---|---|---|---|---|---|
| Feb 19 | 3,386 | +1 | 1.04 | +1.04 | Long $10.4B |
| Feb 26 | 3,116 | +1/3 | 0.75 | +0.25 | Sell $7.9B |
| Mar 11 | 2,741 | −1 | 0.33 | −0.33 | Sell $5.8B more, now short |
- Signal first. In the first week the one and three-month returns turned negative, cutting the signal from +1 to +1/3. That alone would have taken exposure to about +0.35.
- Volatility next. Realized volatility rose at the same time, cutting the scalar from 1.04 to 0.75, so exposure fell to +0.25 instead.
- Then both again. By March 11 the 12-month return was negative too, the signal hit −1, and the scalar had fallen to 0.33. The model went from selling a long to building a short.
Two features make this matter beyond one model. First, many funds use similar look-backs, so their triggers cluster. Second, the selling arrives when liquidity is thinnest. The IMF made the general point after the February 2018 volatility shock, writing in its blog that funds using trading algorithms are apparently more likely to react to a spike in volatility by selling assets. The same mechanism drives volatility control funds, which is why strategists often add the two together.
How do you track CTA positioning with the COT report?
You track it with the CFTC’s Commitments of Traders (COT) report, specifically the Traders in Financial Futures (TFF) version, which shows the net position of the Leveraged Funds category in contracts like the E-mini S&P 500, Treasury futures and currencies.
The CFTC publishes the COT generally each Friday at 3:30 pm Eastern, using positions as of the preceding Tuesday. It comes in three formats: Legacy (commercial vs non-commercial), Disaggregated (producers, swap dealers, managed money and others) and TFF for financial futures. The CFTC’s explanatory notes set out the TFF categories:
- Dealer/Intermediary. The sell side: large banks and dealers that design and sell financial products and tend to run matched books.
- Asset Manager/Institutional. Pension funds, endowments, insurance companies, mutual funds and managers whose clients are mostly institutional.
- Leveraged Funds. Typically hedge funds and money managers, including registered CTAs and commodity pool operators, with outright or arbitrage strategies.
- Other Reportables. Other buy-side traders above the reporting level.
Read it with three limits in mind:
- It is stale. A Tuesday snapshot published Friday misses everything that happened in a fast week.
- It is mixed. Leveraged Funds include relative-value funds that are short index futures as a hedge against long stock baskets, and basis traders in Treasuries (see the basis trade guide). A big net short is not automatically a bearish CTA bet.
- It covers only large traders in markets where at least 20 traders hold reportable positions, as the CFTC notes, and it shows futures and options, not swaps or ETFs.
The useful signal is the change and the extreme. A Leveraged Funds net position at a multi-year high or low, moving the same way as price, is consistent with trend followers being fully positioned. Compare it with your trend model rather than reading it alone.
How reliable are bank CTA estimates?
Bank CTA estimates are model outputs, not disclosed positions, and you should treat them as a map of where pressure could build, not as fact. Research desks at firms such as Goldman Sachs and Nomura build their own trend replicas and publish implied exposure, trigger levels and expected flows under scenarios.
They are useful because the underlying rules are simple. They are fragile because every input is a choice: which look-backs, which volatility window, what assumed assets under management, which markets. Two desks can publish very different numbers for the same day. When you read “CTAs will sell $X billion if the S&P falls below Y”, translate it to: our replica, with our asset estimate, sells that much below that level.
Pay attention to CTA estimates when three things line up: the model is near maximum long, the nearest trigger is within a few percent, and realized volatility is rising. That combination is when systematic selling can turn a dip into a slide. When the model is already short, most of the selling has happened.
When do CTAs buy back?
CTAs buy back slowly, and later than you might expect, because their signals need prices to rise above old levels and their volatility scalars need realized volatility to fall. After a crash both take weeks.
In our model the S&P 500 rose from 2,398 on March 18, 2020 to 2,750 on April 8, a gain of about 15%, yet exposure stayed near −0.17 to −0.20. The signal remained −1 because the index was still below where it had been one, three and twelve months earlier, and high volatility kept the short small. The same pattern showed in 2022, when rallies in March and April only lifted the signal back to +1/3.
That lag is why strategists describe CTAs as a source of buying in the later stage of a recovery: once volatility settles and the look-backs turn up, the model rebuilds longs in steps, sometimes into strength. It is also why short-covering rallies can be sharp. Trend followers who are short buy as prices pass their triggers on the way up.
How should you use CTA positioning in your own decisions?
Use CTA positioning as a context signal for timing and sizing hedges, not as a directional forecast. It tells you where the market’s mechanical pressure points are, not where fundamentals will take it.
Treating a published trigger level as a floor or a ceiling. Trend models do not defend levels; they react to them, and real CTAs blend many signals, so the actual flip is a zone, not a line. Selling puts or buying stock because “CTAs are max long” or “CTAs are short, so it’s the bottom” is a bet on a model of a model.
Practical uses that hold up:
- Hedge timing. If trend models are full and a trigger is close, protection bought before the break is usually cheaper than after it, when implied volatility has jumped. See VIX volatility regimes.
- Strike selection. Trigger zones are reasonable places to think about put strikes or collar floors, because breaks through them can be fast.
- Patience after crashes. When systematic sellers are already short and volatility is extreme, much of the forced selling is done. That is a time to re-sell premium carefully, not to add downside protection at peak prices.
- Crowding checks. Combine CTA estimates with hedge fund crowding data; when both point one way, exits can be violent. See crowded trades.
What does CTA positioning mean if you hold concentrated tech stock?
If you hold a large position in one mega-cap tech stock, CTA positioning matters because index futures selling reaches the biggest index weights first. When trend followers sell S&P 500 and Nasdaq-100 futures, the dealers on the other side hedge by selling the underlying stocks, and the largest names carry the largest weights.
That gives an executive or long-tenured tech employee three practical takeaways:
- Your stock can fall for reasons unrelated to your company. A mechanical index unwind is a market-risk event, and it is exactly the kind of drawdown a collar or index put spread is designed for.
- Covered-call strikes deserve a second look near triggers. When the model is near maximum long and close to its first trigger, consider pairing calls with protection; when the model is short and volatility is high, premium is rich and calls can be written further out.
- Don’t time your diversification plan off flow estimates. Scheduled sales under a 10b5-1 plan should follow your tax and concentration plan, not a strategist’s trigger map.
This is the hedge-fund-style lens Proflex applies to concentrated positions: separate company risk from market-flow risk, and use options to cover the part you can’t diversify away quickly.
What is a weekly checklist for reading CTA positioning?
- Update a simple trend model on the index you care about: signs of the 1, 3 and 12-month returns and 3-month realized volatility.
- Write down the trigger ladder: the closes from one, three and twelve months ago, and how far each is from today.
- Check realized volatility against your target; rising volatility means smaller positions even with an unchanged signal.
- Read Friday’s TFF report for E-mini S&P 500 and Nasdaq-100 Leveraged Funds, focusing on the weekly change and multi-year extremes.
- Compare with bank estimates if you have them, and note where they disagree with your model.
- Map hedges to the ladder: put or collar strikes near the zones, call strikes adjusted to the volatility regime.
- Log what happened after each trigger break so you learn how much weight the signal deserves.
Frequently asked questions
What are CTA trigger levels?
CTA trigger levels are estimated index prices at which trend-following models would change their position, usually because the price crosses a moving average or falls below where it stood one, three or twelve months earlier. They are model estimates, not disclosed orders, and they shift every day as the look-back window rolls forward.
Why does CTA selling accelerate in a selloff?
Two things cut a trend model’s position at once. The signal weakens as more look-back windows turn negative, and rising volatility shrinks position size because exposure is scaled by target volatility divided by realized volatility. In our S&P 500 model both effects took exposure from +1.04 to −0.33 of capital in 15 trading days in 2020.
How do you see CTA positioning in the COT report?
Use the CFTC’s Traders in Financial Futures report and look at the Leveraged Funds category for contracts such as the E-mini S&P 500. The CFTC says Leveraged Funds are typically hedge funds and money managers, including registered CTAs and commodity pool operators, so the number mixes trend followers with other hedge fund strategies.
When is the COT report released?
The CFTC generally publishes the Commitments of Traders report each Friday at 3:30 pm Eastern, showing positions as of the preceding Tuesday. Holidays can delay it. By the time you read it, three trading days have passed.
Are Goldman Sachs or Nomura CTA estimates real positions?
No. Bank research desks build their own trend models and publish what those models imply about CTA exposure and trigger levels. They can be useful, but they are estimates. CTAs do not disclose daily positions, and different desks can publish very different numbers for the same day.
What is CTA deleveraging?
CTA deleveraging is the reduction of trend-following funds’ gross exposure, usually because volatility has risen and position sizes are scaled to a volatility target, or because trend signals have weakened. It can happen even when the direction of the trend has not changed.
Do CTAs cause market crashes?
They can add to a move but rarely start one. Trend models react after prices have already fallen, so their selling tends to accelerate an existing decline. Once they are short and volatility is high, there is less left to sell, and their later buying can fuel rebounds.
Key takeaways
- CTA positioning is estimated from price, not disclosed: trend models are simple enough to reverse-engineer roughly.
- Trigger levels are just old prices, such as where the index stood 1, 3 or 12 months ago, or a moving average.
- Selling accelerates because two multipliers move together: weaker signals and higher volatility both shrink exposure.
- Our simple S&P 500 model swung from +1.04 to −0.33 of capital between February 19 and March 11, 2020.
- The CFTC’s weekly COT report is the only official data, three days stale and mixing CTAs with other leveraged funds.
- Bank CTA estimates are models, not positions; use them as a map of where pressure could build, not as a forecast.
- If you hold concentrated stock, know where systematic sellers sit before you set hedge strikes or call strikes.
Sources and method
- CFTC, Commitments of Traders — Friday 3:30 pm ET release, Tuesday positions, report types
- CFTC, Traders in Financial Futures Explanatory Notes — Leveraged Funds definition, 20-trader threshold, sell side vs buy side
- National Futures Association, CTA Registration — what a commodity trading advisor is
- Moskowitz, Ooi & Pedersen, Time Series Momentum, Journal of Financial Economics (2012), via AQR — 1 to 12-month look-backs, 58 futures markets
- IMF Blog, Volatility Strikes Back (May 3, 2018) — algorithmic funds more likely to sell on volatility spikes
- S&P 500 via FRED (SP500) — daily closes for the trend model, Oct 2016–Oct 6, 2026
Charts 1 and 2, the stat strip and the worked example use a Proflex trend model on FRED SP500 daily closes (Oct 7, 2016 to Oct 6, 2026). Signal = average of the signs of the 21, 63 and 252-trading-day returns (+1, +1/3, −1/3 or −1). Exposure = signal × min(1.5, 10% ÷ 63-day realized volatility, annualized). The first year is the look-back, so the series starts in October 2017. It illustrates how trend followers behave; it is not an estimate of real CTA positions or assets. The dollar figures in the worked example are hypothetical. Chart 3 follows the CFTC release schedule.