iM-DrawdownShield is a rules-based market timer for the S&P 500. It holds the SPDR S&P 500 ETF (SPY) most of the time and moves into a defensive asset when its indicators show that the market’s trend has turned down across all time frames. It is designed for investors who want to stay in the stock market but cannot afford a 50% drawdown, such as those near or in retirement. Every step is fixed and published in advance; nothing is adjusted by hand from day to day.
The data
The model uses only SPY’s daily closing prices, adjusted for dividends and splits, and its daily trading volume. No economic data, sentiment surveys or other markets enter the decision to be in or out of SPY. The defensive assets are rated with the same method from their own prices and volume.
The bull–bear indicator
The core of the model is a bull–bear indicator calculated for three time frames: short term (about one month of trading), medium term (about three months) and long term (about one year). Each time frame produces one value on a scale from −10 (strongly bearish) to +10 (strongly bullish).
Each value combines four independent measures of the market:
- Momentum – how far the price has moved over the period, measured against its normal day-to-day volatility, so that a move in a calm market counts for more than the same move in a turbulent one.
- Trend quality – how steadily the price has moved in one direction, as opposed to an erratic path that happens to end higher or lower.
- Fibonacci position – where the current price sits within its recent high–low range, graded at the classic Fibonacci retracement levels (23.6%, 38.2%, 50%, 61.8% and 78.6%).
- Volume confirmation – whether trading volume is heavier on up days or on down days, showing whether buyers or sellers are in control.
The four measures are combined, smoothed to remove day-to-day noise, and scaled to the −10 to +10 range. The values are published to two decimals and read in five bands:
| Value | Reading |
|---|---|
| +5.5 to +10 | Strongly bullish |
| +2.5 to +5.5 | Bullish |
| −2.5 to +2.5 | Neutral |
| −5.5 to −2.5 | Bearish |
| −10 to −5.5 | Strongly bearish |
The buy–sell value
A second set of three values measures the direction of each bull–bear line: whether it is rising, flat or falling, and how fast. This buy–sell value is also scaled from −10 to +10. It reacts earlier than the bull–bear value itself, because a line can still be bullish while already turning down.
The switching rule
The decision is taken once a day, after the close:
- Leave SPY when all three buy–sell values are falling by more than a set margin. The margin depends on how bullish each time frame still is: the more bullish the market, the stronger the decline required before the model sells. A healthy bull market is therefore not abandoned at the first dip.
- Return to SPY when all three buy–sell values are rising by more than a set margin. Here the margin depends on how bearish each time frame still is: the deeper the bear market, the stronger the recovery required before the model buys back.
- Otherwise the model keeps its current position.
Requiring all three time frames to agree is what keeps trading infrequent. Since 1995 the model has switched about twice a year and has been invested in SPY about 84% of the time.
Where the money goes when out of SPY
Two choices of defensive asset are published:
- Cash – 1–3 month US Treasury bills (BIL).
- Best ETF – on the day the model leaves SPY, the three Treasury ETFs SHY (1–3 years), IEF (7–10 years) and TLT (20+ years) are rated with the same bull–bear method. The one with the strongest positive medium- and long-term rating is bought and held until the model returns to SPY. If none of the three is rated bullish, the money goes into T-bills.
The Best ETF is chosen once, at the exit, and is not changed during the defensive period. Testing showed that switching between bond ETFs while out of the market added trades without adding return.
The switching price: knowing in advance
Every evening the model calculates the SPY closing price that would trigger a switch on the next trading day. If SPY closes beyond that price, the model switches at that close. Because the price is known the evening before, members can place their trade near the close of the same day, rather than a day late.
Most of the time the switching price is far away. The members’ page shows it as follows:
- more than 10% from the latest close: “no signal within ±10%”;
- between 2.5% and 10%: shown for information;
- within 2.5%: highlighted, and a text-message alert is sent.
Alerts
When the switching price comes within 2.5% of SPY’s close, members receive an evening text message naming the trade and the price, for example “Sell SPY, buy SHY near close if SPY < $x”. If the next day ends without a switch and no new alert follows, a stand-down message is sent. Since 1995 this has averaged about seven messages a year.
On rare days the market moves so sharply that the model switches without a prior alert. In that case the trade is executed at the next day’s close, and the published results include this delay. With the current settings this has happened once since 1995, on August 4, 2011.
How the model was built and tested
The settings were developed on data from 2002 to 2019 only. Two periods were held back to test them: 1995 to 2001 (the late-1990s bull market and the dot-com crash) and 2020 to 2026 (the COVID crash, the 2022 inflation bear market and the 2025 tariff shock). The model performed in the held-back periods much as in the development period, reducing the deepest falls by roughly half or more, with returns similar to or higher than those of SPY.
Each setting was chosen in the middle of the range over which the trading dates do not change, so that small differences in a setting do not change the results.
The defensive ETFs did not all exist in 1995. Before each ETF’s launch its history is simulated from Treasury yields, and the simulation was checked against the actual ETF once it began trading.
What the model does not do
- It does not forecast the market. It reacts to trends that are already under way and accepts some loss at the start of each decline.
- It will lag SPY in strong bull markets that contain sharp but short corrections, as in 1998, 2019 and 2023.
- It is designed and tested for SPY only; the rule is not claimed to work for individual stocks or other funds.
- It produces false alarms: about four in ten alert periods end without a trade.
Disclosure
All performance figures are hypothetical results of applying the published rules to historical data. Prices of the defensive assets before each ETF’s launch are simulated. The figures do not represent actual trading and do not include costs, slippage or taxes. Past performance, actual or simulated, does not guarantee future results. iM-DrawdownShield is a mathematical model; it is not investment advice.