SMA stands for simple moving average.
Read from the start: chapter 1 The RuleLeveraged ETFs can lose a large part of their value in a short time.
Getting started
For the first weeks, before you buy anything.
The 200-day line (SMA 200) is the average of the last 200 closing levels of an index. If the index closes above it, you are invested in the leveraged product; if it closes below, you get out.
What is meant is a simple moving average over 200 trading days. It smooths the price movement and moves more slowly than the index itself.
Trend following checks once a day at the close which side of the line the index is on. A trade happens only when it changes sides; in between there is nothing to do.
The line is computed on the unleveraged index, not on the price of the leveraged product.
200 days is the common value for the S&P 500. For other indices this page computes with other lengths, 220 trading days for the Nasdaq-100 and 255 for the MSCI World; the principle is the same. Which length, which buffer and which signal currency your version uses is on the rule card.
A leveraged ETF moves each day by a multiple of its index. Because the leverage refers to the day, the outcome over months also depends on how turbulent the path was.
A leveraged product moves each day by a multiple of its market's move, up as well as down. Which multiple applies is stated with the product and on the rule card.
The word daily matters: the leverage applies to the daily move, not to a year. Over longer periods the outcome therefore also depends on how calm or turbulent the path was.
There is no minimum amount. With small amounts, fixed order fees and spreads on every switch eat up the advantage the rule is meant to bring.
The rule has no minimum amount. What weighs on small amounts are the fixed costs per order and the bid-ask spread, because you pay both on every switch regardless of the size of the position.
With small amounts that can eat up the very difference the rule is meant to produce. What an order costs at your broker, and whether it trades fractions, decides more here than the strategy does.
The rule allows entry as long as the index is above the line, however far. The price: even a pullback to the line is then a loss, including for a lump-sum investment.
The rule has no entry condition. As long as the series is above the line, entering is consistent with the rule, however far above it is.
The flip side belongs with it: entering far above the line leaves no cushion of gains, and even the pullback to the line is felt as a loss. Both sides are part of the same rule; which one counts for you is your call.
Two ways are common for not entering with the whole amount far above the line: waiting for the next cross from below, and spreading the entry over a few months. The backtest on this page reproduces neither of them exactly.
Then the rule first shows you its current state: above the line, holding is consistent with the rule; below it, the rule would be out. Whether you follow it from today or from the next signal is your decision.
Then the rule first tells you which state it is in. While the market is above the line, holding is consistent with the rule; while it is below, the rule would be out.
Whether you follow a rule from today or only from the next signal is your decision, not a statement of the rule. The curves on this page start at the beginning of the period you set, not at your purchase date.
Savings plans and inflows
When new money arrives regularly.
If it keeps running in the leveraged product, you are buying back exactly what the rule has sold. Two ways fit the rule: pausing the instalments and accumulating them, or redirecting them into the fallback block.
If the plan keeps running in the leveraged product, you are buying back exactly what the rule just sold. For that part of your money this is simply holding.
Two ways fit the rule: pausing the instalments and accumulating them, or redirecting them into the fallback block. Which of the two does better has no reliable answer.
Above the line the rule is invested, so both fit: investing the amount at once or spreading it over a few months. Waiting for a pullback would be a second rule.
The rule says nothing about it. It knows two states, invested or out, and above the line the state is invested. An amount arriving on top does not change that.
Two ways stay consistent with the rule: put the amount into your version in one go, or spread it as a savings plan over a few months. Neither produces additional signals.
Waiting for a pullback down to the line, by contrast, is a second rule beside the first one. There is no backtest for it on this page, and the curves you see here do not apply to it.
Not in the leveraged product; money market funds, instant-access savings or short-dated government bonds are usual. This page credits the time out with the money market rate.
The rule only asks that it does not sit in the leveraged product. Where it goes instead belongs to implementation, not to the rule.
The usual candidates are a money market fund, instant-access savings or a short-dated government bond. None of them is risk free, and what comes out depends on the interest rate level of the time.
A common variant switches below the line into the unleveraged index instead of into money market or instant-access savings. This page does not compute it, and the taxable event at the switch remains.
In the chart this side lives in the interest assumption: it credits the time out of the market with the money market rate of the product's currency; negative rates count as zero. Interest is taxed, and what that means for you is in the chapter Implementation and Taxes.
Day to day
What happens between the signals.
The rule applies to daily closing levels, so once per trading day. Most of the time there is nothing to do.
The rule is defined on daily closing levels, so yes, once per trading day. In most cases there is nothing to do, and LETF.io notifies you when a signal appears.
Looking only once a week means trading a different rule from the one on file. That is possible, but then the backtest of the rule on file no longer applies to you.
The backtest on this page trades at the close of the signal day. Trading on the next trading day gets you a different price; the difference is usually small, but after a price gap it can be large.
For a US index the closing level is usually fixed at 22:00 German time, and at 21:00 in the weeks between the clock changes in the US and in Europe. Some venues are still trading then, often with a wider spread; most investors trade the next morning.
If the index is back on the other side of the line the next morning, the signal from the close still applies. The rule knows only closing levels, no levels during the session.
What matters is that you do it the same way every time. Trading sometimes at once and sometimes later, depending on mood, adds a second rule to the first.
Trading a signal from two days ago at today's price is your own decision, not an execution of the rule. Writing the decision down keeps you from later booking it as a result of the rule.
Trading a signal from two days ago at today's price is a decision of its own, not an execution of the rule. The backtest you have seen does not cover it.
In practice there is no way around deciding and writing the decision down. What matters is that you do not later book it as a result of the rule.
Only the closing level. If the price touches the line during the session, that is not yet a signal, and futures trading overnight neither create one nor take one back.
The rule knows only the closing level. While the trading day is still running there is no signal, even if the price touched the line hours ago.
Futures that keep trading overnight are not a price as far as the rule is concerned. They can show you where the mood is going, but they do not create a signal and they do not take one back.
In practice that means one look after the close is enough. For a US index that moment falls in the European evening, and the notification from LETF.io comes after it.
A notification is created after every close in which the index changes sides of the line. A price alert in a charting program often computes on a different series.
The notification comes after the close of the index, so for a US index late in the European evening. It names the side of the line and the distance in percent.
An alert you set on the line yourself in a charting program often triggers during the session and frequently computes on the price index without dividends. Why that leads to different days is under the question on the difference from TradingView.
That is not a signal but a duty of your broker: for leveraged products it reports every loss in value of 10 percent against the initial value and then every further step of ten.
That is not a signal from your rule but a duty of your broker. For leveraged products it has to notify you as soon as the initial value of the instrument has fallen by 10 percent, and after that at every further loss in steps of ten.
What counts is the value the position started at, not the level of your last statement. Whether a position that has recovered is reported again when it falls below the threshold once more is handled differently from broker to broker. The basis is Article 62(2) of Delegated Regulation (EU) 2017/565; the purpose is transparency.
With a leveraged holding these notices therefore arrive more often than with an unleveraged one, and they come especially when markets are turbulent. What your rule does is in the rule, not in the letter.
No, that is a different rule: a stop loss hangs on the price of your product instead of the index and triggers during the session instead of at the close.
No, a stop loss order is a different rule. It hangs on the price of your product instead of the index series, often in a different currency, and it triggers during the session instead of at the close.
The index signal cannot be cleanly translated into a standing order at a broker. What remains is looking every day and the question of who acts for you when you cannot.
Nobody, unless you prepare it. That takes a power of attorney for your account and the rule in writing, so that a person without prior knowledge can carry it out.
A rule that checks daily assumes a person who looks. Ill, travelling without a connection, in hospital: there are weeks in which nobody does.
Two things belong with that. A power of attorney for your account, so that somebody is allowed to act at all. And the rule in writing, in a form a person without prior knowledge can carry out: which value, which condition, which order.
If there is nobody you want to let in on it, the number of signals of your version counts among the practical questions rather than the marginal ones.
Products and leverage
What you carry out the rule with.
The index your product tracks, so for an MSCI World 2x the MSCI World, not the S&P 500. A foreign signal series crosses the line on different days and is effectively a different rule.
Each version belongs to exactly one product, and by default the signal series is the unleveraged index of that product. That is why the same strategy exists several times, once per product.
A product on a similar index is not the same thing. The line is then crossed on different days, and you are effectively trading a different rule.
Two. Each version belongs to one product and computes its signal on that product's index; the two lines cross on different days.
A common signal, say only from the broader index, would be a third rule, for which this page has no backtest.
LETF.io carries each version separately, with its own line and its own notification.
No. With leverage the volatility drag in turbulent phases grows more than proportionally, and the filter does not protect between the high and the signal, where higher leverage bites hardest.
No, the relationship runs the other way. The higher the leverage, the stronger the volatility drag in turbulent phases, and the more expensive every phase the filter does not catch.
The filter also does not reduce the risk between the high and the signal. That is precisely where higher leverage bites hardest.
For a sense of scale, at an assumed index volatility of 18 percent a year: the extra drag comes to about 3 percentage points a year at 2x and just under 10 at 3x, before the cost of financing the leverage.
Its benefit shrinks faster than the leverage, its costs stay. The extra drag grows with L(L−1)σ2/2 and vanishes at leverage 1; spreads, taxes and missed recoveries do not.
The reason for the filter grows weaker, but it does not disappear. The extra drag against L times the market grows with L(L−1)σ2/2 and not with the square of leverage: it falls more than proportionally when leverage comes down, and at leverage 1 it is gone entirely.
The costs of the filter, on the other hand, stay as they are. Bid-ask spreads, tax on every exit and missed recoveries fall due regardless of leverage.
At low overall leverage the balance therefore shifts against the filter. Where exactly the line runs depends on your costs and your tax situation, and this page does not know those.
Yes, but its main reason falls away: without leverage no additional volatility drag arises. What remains is limiting deep declines, at the same costs.
You can, but its main reason falls away. The additional volatility drag it is meant to contain does not arise without leverage.
What remains is the limiting of deep declines, and you keep paying spreads, taxes and missed recoveries for it. Without leverage declines are shallower, so limiting them weighs less.
Because there is more between the two than the fee: the financing of the leverage inside the price, the effect of the daily leverage, and swap and dividend costs. In calm trends the product can also come out ahead.
Over longer periods every product deviates, downwards in turbulent phases and upwards as well in calm trends. There is more between the two than the fee. A leveraged product borrows the part it holds on top, and the cost of that does not sit in the expense ratio but in the price.
On top of that comes the volatility drag from the daily leverage, and the distance every product has from its index, for example from swap costs and from taxes on dividends.
That distance is also the reason why signals are computed on the index and not on the price of the product.
The financing alone costs roughly the leverage minus one times the short rate: for a 3x product in dollars about 7.8 percent a year (2 times 3.9 percent SOFR), for a 2x product in euro about 2.4 percent (€STR), as of November 2026, in each case plus the swap counterparty's markup.
That depends on the currency in which the product's leveraged index is computed and financed. Whether the index currency and the product currency differ says little about it.
That depends on the product, more precisely on the currency in which it finances its leverage. If it finances in euro, the leverage acts on the exchange rate move as well. If it finances in dollars, you carry the exchange rate move unleveraged.
You can read it off the name of the leveraged index in the key information document. A euro product on a leveraged index of a US market computed in euro leverages the exchange rate too, even though index and product currency are the same.
The currency is therefore one of the reasons a product moves away from index times leverage.
That would be a second rule. Phases below the line are mostly short, a short product carries a larger volatility drag than a long product, and this page does not compute the combination.
That would be a second rule, not this one. The rule on this page knows two states, leveraged and invested or out; a short holding is a third.
Two points belong with it. Phases below the line are mostly short, and the line often keeps rising through them out of its inertia. In these short phases a short holding therefore rarely earns anything, and a short product carries a larger volatility drag than a long product with the same leverage.
For the combination of filter and short holding this page shows no backtest. What stands here applies to the rule that stands here.
That raises the overall leverage and is a different rule. The loan keeps running when the rule exits, until you repay it, and no curve here includes interest or margin call risks.
A securities loan is secured by the account. If its value falls, the bank can demand more collateral or sell, regardless of what the rule says at the time.
What has been said about more leverage applies twice over here: the volatility drag grows, and between the high and the signal the whole holding is invested.
The provider sells and pays out, and that realises gains or losses, whatever the rule says. You then need a new product and therefore a new version.
It happens: providers discontinue leveraged products. With a fund the assets are sold and paid out; with a debt security it is redeemed under the procedure laid down for it.
Either way that realises gains or losses, regardless of what the rule says at the time. For the strategy it means you need a new product and therefore a new version, because by default the signal series depends on the index of that product.
Parameters and the calculation
When you recompute or compare numbers.
Usually because of one of four points: a different series (price instead of total return index), trading days instead of calendar days, a different currency, or a level before the close.
Usually it is one of four points. First the series: we compute on the index series that belongs to your product, and in the variant the product tracks; many programs show the price index without dividends.
Second the counting, because our window counts trading days and not calendar days. Third the currency the signal is computed in. Fourth the moment: the closing level counts, not the state in the afternoon.
Which four settings your version uses is on the rule card. That is where the comparison sensibly starts, not at the line itself; mixing up the counting is the most frequent case (clarification on it).
A fifth, rarer point is the trading calendar: MSCI series also have a value on US holidays, and then 200 days span a different period.
There is no right length, only a range in which results come out similar. The length that looks strongest in hindsight is usually fitted to that exact price history.
There is no right length, but a range of usable lengths in which results come out similar. The length that looks strongest in a backtest is usually fitted to that exact price history.
The range shows you how sensitive the rule is to the setting. We offer no parameter search: search long enough and every price history yields a setting that looks magnificent.
Changing the length after every false signal means changing the rule each time in the moment when the old one has just been uncomfortable.
A buffer requires the index to cross the line by a fixed distance before a trade happens. That filters out part of the false signals and costs ground on every real signal.
With a 3 percent buffer the rule sells only when the index closes 3 percent below the line, and buys again only when it closes 3 percent above it. If it oscillates just around the line, the holding stays as it is.
The price: exit and entry come later. Whether that pays off depends on how many sideways phases a period contains, and that changes from decade to decade.
A buffer is a parameter like the window length, and the same applies: the value that looks strongest in hindsight is fitted to that exact price history. Your version shows which buffer it computes with.
For a sense of scale on the S&P 500 with net dividends and 200 days, October 2000 to October 2026: without a buffer 161 signals, with a 1 percent buffer 73, with 3 percent 23. In the sideways year 2015/16 the number fell from 18 to 10 with 1 percent and to 2 with 3 percent.
Through the ratio of about 252 trading days to 365 calendar days a year. More important is whether the foreign figure was meant in trading days at all; if that is not stated, the comparison is uncertain.
Through the ratio of trading days to calendar days. A year has just under a third fewer trading days than calendar days; a figure in calendar days therefore becomes noticeably smaller when you translate it into trading days, and a figure in trading days correspondingly larger.
More important than the exact conversion is the question of which unit the foreign figure meant in the first place. Where it is not stated, the comparison is uncertain.
As a rule of thumb: 200 trading days are about 290 calendar days, and 200 calendar days are about 138 trading days.
That is a different rule. The EMA reacts earlier and more often reports a move that was none; all numbers on this page apply to the simple average.
That is a different rule. An exponential average gives more weight to the most recent prices, so it reacts earlier and in exchange reports more moves that were none.
Everything computed and shown on this page holds for the simple average. A version with a different average would have different signal days and therefore different backtests as well.
As long as LETF.io does not carry the exponential average as a version of its own, there are no numbers for it here.
Because the product price contains fees, leverage, currency and trading hours. The statement about the trend should come from the market; two products on the same index therefore get the same signal.
Because the product price contains fees, leverage, currency and trading hours. The statement about the trend should come from the market, not from the quirks of one product.
In practice that means two products on the same index get the same signal, even though their prices run differently.
If the series is computed in euro, dollar movements sit inside the signal; in the index currency the exchange rate stays out. It does not leave your account in either case.
That depends on the signal currency of your version. If the series is computed in euro, the dollar's movement sits inside the line, and you get signals that come partly from the currency.
If it is computed in the index currency, the exchange rate stays out of the signal. It does not leave your account, though: your product still depends on the same market and the same currency.
Backtests and risk
What the curves say and what can go wrong.
Deep. Between the last high and the signal day you are fully invested, and with leverage that stretch is correspondingly severe.
Large. Between the last high and the signal day you are fully invested, and with a leveraged product that stretch is correspondingly severe. Why that is so is in the chapter Typical Mistakes.
In sideways phases the index often crosses the line several times in quick succession. Every false signal costs spread, often tax, and some price, because you sell lower and buy back higher.
False signals belong to the rule; without them there would be no timely exits either. On the S&P 500 with 200 days there were 161 signals from October 2000 to October 2026, about 6 a year; in the sideways year from July 2015 to July 2016 there were 18. How many there were in your version is shown in the chapter Backtest with the signals per year.
What a false signal costs depends on three things: the gap between the selling and the buyback price, your trading costs, and the tax on the realised gain. The last item is missing from most backtests.
A buffer or a longer window lowers the number of false signals and in exchange makes every real signal later.
Arithmetically, at a daily index loss equal to the reciprocal of the leverage, which never happened with these products. The real path to very large losses is many bad days in a row.
Arithmetically total loss would be reached if the market fell by the reciprocal of the leverage in a single day. For the products this strategy uses such a daily move has never occurred. The worst day of the S&P 500 since 2000 was about minus 12 percent, the 1987 crash minus 20.5 percent; at minus 20 percent US trading today halts until the close.
The real path to a very large loss is a different one: many bad days in a row. Over months, leveraged holdings have shown declines that cost the largest part of the amount committed. That belongs to this strategy.
What remains is the risk overnight and over the weekend. The rule does not check during that time, and a market can move far between the close and the opening.
Nobody knows in advance. The idea does not come from a parameter search and limited deep declines in backtests; the exact window length, by contrast, was chosen on the past.
What speaks for the idea is its reason: the volatility drag of a leveraged product is largest in turbulent, falling markets, and in such phases the index is mostly below the line.
What speaks against too much confidence is that every backtest knows only one price history. That is why this page shows ranges of window lengths instead of one best number, and the chapter Criticism collects the objections.
They show what the rule would have done, not what it will do. The backtest reproduces the time before the products with a leverage model; its limits are overlapping windows and the choice of parameters.
That depends on which chart you are reading. The chart in the chapter The Rule shows the recent years on the price series itself; the backtest computes on the long index history and reproduces the time before that with a leverage model.
Both show what the rule would have done, not what it will do. What that means for their weight is in the chapter Backtest: overlapping windows, parameters chosen against the past, a single interest rate environment.
Not reliably; that depends on the period. Two comparisons make sense: with the held leveraged product (did the rule help?) and with the unleveraged market (did the leverage help?).
In the chart the reference is holding the same product, that is the leveraged holding without the rule. That comparison answers the question of whether the rule achieved anything.
The second sensible reference is the unleveraged market. It answers the other question, namely whether the leverage achieved anything.
The two belong together. A rule can beat holding and still sit behind the unleveraged market, and it can sit ahead of the unleveraged market without the rule itself having contributed to that.
One year says little. A meaningful period contains at least one long decline and several sideways years in which the filter only costs.
A single year says little. The filter bites in long declines, and such phases are rare; in between lie years in which it only costs.
A period that says something about the rule contains both: at least one long decline and several sideways years. The distribution in the chapter Backtest shows exactly that: the spread across the rolling periods.
Anyone who stops after a bad year leaves the rule and afterwards no longer knows what they are actually testing.
Taxes
What your tax case does to the rule. Tax information as of November 2026; not tax advice.
Every exit realises gains or losses, and tax paid out no longer compounds. Few signals are therefore more economical in tax terms than many.
That depends on your region and is in the chapter Implementation and Taxes of this page. The core is the same everywhere: every exit realises gains or losses, and what you pay in tax no longer compounds.
A strategy with few signals is therefore more economical in this respect than one with many. In a backtest without tax you cannot see that difference.
Realised losses from fund units go into your bank's general loss pot and are offset against later investment income. What income remains after that is tax free up to the saver's allowance of 1,000 euros a year with an exemption order.
Losses are therefore not lost, but they only take effect once income arises again. If they are left over at the end of the year, the bank carries them forward into the next year, unless you apply for a loss certificate by 15 December (§ 43a Abs. 3 EStG).
For products that count as equity funds, the partial exemption also applies to losses: 30 percent of them do not count for tax, just as 30 percent of a gain is tax free (§ 20 InvStG).
The saver's allowance is 1,000 euros, 2,000 euros for joint assessment (§ 20 Abs. 9 EStG). As of November 2026, not tax advice; how it all adds up in your case is a matter for tax advice.
The advance lump sum hits the fund, not the rule: whoever simply holds pays it as well. The rule brings the rest of the tax forward.
The advance lump sum does not hit the rule but the fund. It falls due in every year in which a fund has risen in value and distributed little, even without you selling anything. Whoever simply holds pays it as well (§ 18 InvStG).
For the comparison between rule and holding that means: part of the tax falls on both sides. What the rule adds is the earlier moment for the rest.
Two qualifications belong with it: it only applies when the base rate is positive (in 2021 and 2022 it was not), and it is credited at the later sale, so you do not pay it twice (§ 19 Abs. 1 InvStG). Selling before the end of the year means paying none for that year.
Losses the rule realises land in the same pot for investment income and reduce the tax on later income (§ 20 EStG). As of November 2026, not tax advice; how it all adds up in your case is a matter for tax advice.
Inside the wrapper the switches trigger no tax for you. Against that stand ongoing fees, often a profit share, no partial exemption, and the issuer risk.
Inside such a wrapper the switches trigger no tax for you, because it is not you who buys and sells. Tax falls due only when you sell the paper itself. That is the point behind the question.
Against it stand three things: an ongoing fee for the wrapper, often a share of the gain, and no partial exemption, because a certificate is not a fund. On top of that it is a claim against the issuer.
This page does not compute such wrappers. What the tax deferral is worth against fees and issuer risk depends on the terms of the individual paper and cannot be said in general.
If you buy the same or a substantially identical security within 30 days before or after a loss sale, the loss is deferred and moves into the cost basis. It takes effect later, not immediately.
If the rule sells at a loss and you buy a substantially identical security within 30 days before or after the sale, the loss is disallowed for now and moves into the cost basis of the new position (26 U.S.C. § 1091). That is exactly the typical path of this rule: a sale after a decline, a repurchase as soon as the series is back above the line.
The loss is not gone by that, it takes effect later. For the calculation it means that a backtest which offsets losses immediately books the advantage in a US taxable account too early.
Inside a tax-advantaged retirement account the question does not arise. But if you buy back there what you sold at a loss in the ordinary account, the loss is lost for good (Rev. Rul. 2008-5). Savings plan purchases in the 30 days before the sale count too.