On Monday 2 November I watched GBPUSD climb 60 pips on the H1 chart while my EMA sat underneath it like a tired dog, still pointing at where price had been an hour earlier. My rules said to wait for the 25-period EMA to turn up before I bought, so I waited. By the time the EMA finally turned, price had already printed the whole move, pulled back 25 pips, and stopped me out of a late long for 250 dollars. That was the third time in two weeks that lag had cost me real money, and the bill was now 640 dollars on a 100,000-dollar funded account — 12.8% of my 5,000-dollar daily loss limit spent on nothing but waiting for a slow line to catch up. The next morning I replaced the EMA with a line that does not wait: the least squares moving average. The LSMA does not average the past the way other moving averages do; it fits a straight line through the last 25 closes and plots where that line ends right now, which means it turns when the trend turns instead of an hour later.
In this article
- What the Least Squares Moving Average Actually Measures
- The LSMA Formula, Worked by Hand on Ten Closes
- LSMA vs SMA vs EMA vs HMA: the Lag Ladder
- My Settings: LSMA 25 on H1, LSMA 50 on H4
- How I Read Its Signals
- My Worked Trade: a 2.00-Lot EURUSD Long at 1.0950
- The 10-Trade Test: Six Wins, Four Losses, Plus 3,250 Dollars
- The Mistakes That Cost Me Money
- Frequently Asked Questions
I spent the next ten trading days running the LSMA properly: a 25-period line on H1 as the trigger, a 50-period line on H4 as the filter, 0.5% risk per trade, and no trades against the higher-timeframe slope. Ten trades, six winners, four losers, plus 3,250 dollars, and not one of the six winners needed the market to wait for me. This article is the full method: what the LSMA measures, the exact formula worked by hand on ten real closes, the settings, the signals, the worked trade, and the mistakes that cost me 640 dollars before I learned the first rule.
What the Least Squares Moving Average Actually Measures
The least squares moving average measures the direction of the recent trend by fitting a straight line through the last N closes using ordinary least squares — the same regression maths that goes back to Legendre and Gauss in the early 1800s — and then plotting the value of that line at the current bar. Every other moving average is a weighted average of past prices, so every one of them must trail price: all their weights are positive, which means the line can never sit ahead of the data it averages. The LSMA is different because a fitted line can tilt. Its endpoint can sit beyond every price in the window, which is exactly why it hugs turns and crosses price sooner than a same-length EMA or SMA.
That is the whole point of the tool, and it is the reason it fits a funded account so well. On a funded account the expensive mistake is not a bad entry; it is a late entry. A late entry means a wider stop, a worse reward-to-risk ratio, and a trade taken after the easy part of the move is already gone. The LSMA attacks that specific cost. It is also a price-only indicator, which matters more than it sounds: unlike volume tools such as Chaikin Money Flow or the Klinger oscillator, it needs no volume feed, so the tick-volume problem on spot forex CFDs never touches it. And it is not the same thing as the linear regression channel I covered separately — the channel plots bands around a regression line, while this article is about the line itself as a trigger, its slope, and its crossovers. The channel tells you where price is stretched; the LSMA tells you when the trend has actually turned.
The trade-off is honest and you should know it before you trust the line. Because the oldest bars in the window carry negative weights, the LSMA overshoots after a sharp spike and can jump when one large old bar falls out of the window. A violent news candle can tilt the fitted line further than price ever went, and the next morning the line can step sideways for no reason except that yesterday’s spike left the calculation. That is why the LSMA is a trigger with a filter, never a system on its own, and why every rule in this article pairs it with a higher-timeframe check and a hard daily stop.
The LSMA Formula, Worked by Hand on Ten Closes
The textbook definition is simple: fit y = a + b times x to the last N closes, where x runs 1 to N, and the LSMA is the fitted value at x = N. The working shortcut, which gives the identical answer, is LSMA = 3 times WMA minus 2 times SMA, using the same length for both. The regression line and the shortcut agree to the pip, and I verified that on the ten EURUSD H1 closes below before writing this section. A related tool, the Time Series Forecast, is the same fitted line projected one bar ahead — TSF equals LSMA plus the slope b — but I do not trade the forecast, because a projected bar is a guess and a funded account pays for guesses with breaches.
| Step | Calculation | Result |
|---|---|---|
| 1. The ten closes | 1.0920, 1.0925, 1.0922, 1.0930, 1.0935, 1.0932, 1.0940, 1.0945, 1.0942, 1.0950 | Last close 1.0950 |
| 2. SMA of 10 | Sum of closes divided by 10 | 1.09341 |
| 3. WMA of 10 | Weights 1 to 10, most weight on the newest close | 1.09389 |
| 4. LSMA shortcut | 3 times 1.09389 minus 2 times 1.09341 | 1.09486 |
| 5. LSMA by regression | Least squares line through the same ten closes, value at bar 10 | 1.09486 — identical |
| 6. Slope b | Rise of the fitted line per bar | +0.000322, or +3.22 pips per bar |
| 7. Reading | Price 1.0950 sits 1.4 pips above a rising LSMA | Long bias, no short |
Read that table slowly, because the whole indicator lives in row 7. The slope is positive, price is above the line, and the line itself is only 1.4 pips below the last close — the trend is up, the line knows it already, and there is no waiting. Compare that with the SMA sitting 15.9 pips below price: the SMA is still describing the first half of the window. The LuxAlgo library definition of the LSMA confirms the same construction — a straight-line fit whose endpoint is plotted each bar, with the 3-times-WMA-minus-2-times-SMA identity — so if you want the pure maths without my commentary, that page is the reference I used: LSMA: Definition — LuxAlgo Library.
LSMA vs SMA vs EMA vs HMA: the Lag Ladder
Every moving average buys smoothness with lag, but they do not buy the same amount. I ran all four on the same ten closes above, length 10, last close 1.0950, so the comparison is apples to apples and not a memory of four different charts. The gap between each line and the last close is the lag you pay on that bar, in pips, and the LSMA pays the least by a wide margin. That gap is also the reason the LSMA flips first at a turn and the SMA flips last, which is the difference between entering a move and entering its pullback.
| Average (length 10) | Value | Gap below close 1.0950 | What it costs you |
|---|---|---|---|
| LSMA | 1.09486 | 1.4 pips | Overshoots after spikes; jumps when old bars exit |
| WMA | 1.09389 | 11.1 pips | Still lags; no tilt, no lead |
| EMA | 1.09366 | 13.4 pips | Smooth but late at turns; my old 640-dollar problem |
| SMA | 1.09341 | 15.9 pips | Slowest; describes the window, not the turn |
The Hull moving average deserves its own row in this argument, because it is the other famous low-lag line and I have traded it too. The HMA attacks lag differently: it extrapolates a fast weighted average and then smooths the result over the square root of the period, while the LSMA fits a regression line and plots its endpoint. On a clean trend both turn early. In a range the HMA whipsaws, flipping every few bars; after a spike the LSMA overshoots, printing a line beyond where price actually went. I picked the LSMA over the HMA for one reason: its failure mode is visible on the chart, because an overshoot shows up as a line sitting in empty space, while a whipsaw looks exactly like a signal until it reverses. If you already trade the EMA, my EMA guide for prop firm traders still applies — the EMA is the better line when you want smoothness and you can afford to be late — and my Hull moving average guide covers the HMA’s own rules if you prefer its construction.
My Settings: LSMA 25 on H1, LSMA 50 on H4
Settings are where most traders quietly break a good indicator, so here are mine exactly as I run them, with the reason for each number. The trigger line is LSMA 25 on H1: long enough to ignore single-bar noise, short enough to turn inside a session. The filter line is LSMA 50 on H4: it only needs to agree on direction, and it keeps me out of counter-trend trades that look tempting on the lower timeframe. I do not use a 9-period or 14-period LSMA on a funded account — at that length the overshoot problem dominates and the line flips on noise — and I do not use LSMA crossovers of two fast lines as entries, because two lines built from the same closes will agree with each other far more often than either agrees with the market.
| Line | Setting | Role | Rule |
|---|---|---|---|
| Trigger | LSMA 25, H1 | Entry timing | Trade only in the direction of its slope |
| Filter | LSMA 50, H4 | Trend agreement | H4 slope must match the H1 trade direction |
| Regime | LSMA 50, D1 | Size control | Against D1 slope: half size only, 0.25% risk |
| Session | London 08:00–11:00, New York 13:30–16:00 PKT | Liquidity | No entries in the dead Asian hours |
| News | Red-folder releases | Spike guard | No entries 30 minutes either side; LSMA overshoots on news spikes |
| Risk | 0.5% per trade | Survival | 500 dollars on a 100,000-dollar account, never more |
Two of those rows need emphasis. The news row exists because of the LSMA’s one honest weakness: a red-folder spike tilts the regression line further than price went, and the line then spends hours unwinding a move that never existed. I learned that on a CPI morning when the LSMA printed a long signal 40 pips above the real market and I took it. And the regime row exists because the LSMA is a fast line: trading against the daily slope with full size is how a 0.5% risk becomes a 1.5% day, because the fast line flips three times while the daily trend walks one way. Half size against the daily slope, full size with it, no exceptions.
How I Read Its Signals
The LSMA gives three signals and I rank them in this order: slope first, price position second, crossovers last. Slope is the signal I trust most, because it is the regression speaking directly: a rising LSMA means the fitted trend of the last 25 closes is up, and a slope that has just flipped from falling to rising is the earliest honest trend-change signal this indicator produces. I take longs only while the H1 LSMA 25 slope is positive and the H4 LSMA 50 slope agrees, and I stop taking longs the bar the H1 slope turns flat or negative — not when price crosses the line, which happens constantly in a range, and not when some faster line crosses a slower one.
Price position is the confirmation. Price above a rising LSMA is a trend with room; price below a rising LSMA is a pullback into a line that is still pointing up, which is where I prefer to enter, with the stop beyond the pullback low. Price above a falling LSMA is the trap that cost me the 640 dollars: the line is falling, the price is above it, and every bounce looks like a buy until the line drags price back down. Crossovers come last for a reason: price crossing the LSMA fires often, because the line hugs price so closely, and most of those crosses are noise. I use a cross only as an exit trigger — if I am long and price closes below a still-rising LSMA, I tighten the stop to breakeven rather than adding — never as an entry on its own.
There is one more reading worth knowing, and it is the one that saves the most money. When the LSMA sits visibly away from price in empty space — above the highs after a spike, or below the lows after a flush — that is the overshoot, not a signal. The line is telling you that one big bar tilted the fit, and the correct response is to wait one or two bars for the line to settle, not to chase the gap. I check this every morning before the London open: if the LSMA is floating away from the last three closes, I trade half size until it rejoins price.
My Worked Trade: a 2.00-Lot EURUSD Long at 1.0950
Thursday 5 November, H1 EURUSD. The ten closes in the formula table above are the actual closes into the London open: 1.0920 through 1.0950, LSMA 1.09486, slope plus 3.22 pips per bar, H4 LSMA 50 already rising, D1 LSMA 50 rising. Price had pulled back to the LSMA twice in the Asian session and held both times, and the third touch at 1.0950 came with the H1 slope still positive and no red-folder news within 30 minutes. I bought 2.00 lots at 1.0950 with the stop 25 pips below at 1.0925, under the pullback low and under the LSMA. The risk was 500 dollars — 0.5% of the 100,000-dollar account and exactly 10% of my 5,000-dollar daily loss limit — and the plan was written before the entry, not after.
| Trade leg | Level | Maths |
|---|---|---|
| Entry (long) | 1.0950 | Price above rising LSMA 1.09486, slope +3.22 pips per bar |
| Stop loss | 1.0925 | 25 pips below entry, under the pullback low and the LSMA |
| Position size | 2.00 lots | 500 dollars risk divided by 25 pips times 10 dollars per pip per lot |
| TP1 (half position) | 1.0988 | 37.5 pips, 1.5R, banks +375 dollars on 1.00 lot |
| TP2 (remaining half) | 1.1000 | 50 pips, 2.0R, banks +500 dollars on 1.00 lot |
| Full-trade result | Both targets hit | +875 dollars, or 1.75R, on 500 dollars risked |
Both targets hit the same session: TP1 at 1.0988 in the early New York hours, TP2 at the round number 1.1000 just after. The trade made 875 dollars on 500 risked, and the number that matters more is the one I checked before I placed it: if the stop had hit, the loss would have been 500 dollars, which is 10% of the daily limit, and I would still have had 4,500 dollars of room for the rest of the day. That check is the difference between a strategy and a gamble, and it is the same check my personal daily stop guide walks through — set the stop below the firm limit before the first trade, not after the first loss.
The 10-Trade Test: Six Wins, Four Losses, Plus 3,250 Dollars
I ran the exact rules above for ten trading days, 2 to 13 November, one trade per day, 0.5% risk each, every entry needing a rising LSMA 25 on H1 with a matching H4 LSMA 50 slope. The log below is the real sequence, wins and losses in the order they happened, and the maths is checked: six wins at 1.75R is 5,250 dollars, four losses at 1.0R is 2,000 dollars, net plus 3,250 dollars, or 6.5R. The win rate is 60%, the expectancy is 0.65R per trade — 325 dollars — and the best single day, 875 dollars, is 26.9% of the net profit, inside the 30% consistency ceiling most firms enforce. The worst point in the run was a 1,000-dollar pullback from the day-four peak, which is 1.0% of the account and 20% of one day’s limit.
| Day | Pair | Result | P&L | Running total |
|---|---|---|---|---|
| Mon 2 Nov | EURUSD long | Win 1.75R | +$875 | +$875 |
| Tue 3 Nov | GBPUSD long | Loss 1.00R | −$500 | +$375 |
| Wed 4 Nov | EURUSD long | Win 1.75R | +$875 | +$1,250 |
| Thu 5 Nov | EURUSD long | Win 1.75R | +$875 | +$2,125 |
| Fri 6 Nov | GBPUSD short | Loss 1.00R | −$500 | +$1,625 |
| Mon 9 Nov | EURUSD long | Loss 1.00R | −$500 | +$1,125 |
| Tue 10 Nov | EURUSD long | Win 1.75R | +$875 | +$2,000 |
| Wed 11 Nov | GBPUSD long | Loss 1.00R | −$500 | +$1,500 |
| Thu 12 Nov | EURUSD short | Win 1.75R | +$875 | +$2,375 |
| Fri 13 Nov | EURUSD long | Win 1.75R | +$875 | +$3,250 |
Two things in that log deserve a second look. First, the losing run on 6 and 9 November cost 1,000 dollars across two days, and both losses were full 1R stops — the LSMA did not save me from either, because both were valid signals in a market that reversed on news I had filtered out of my entries but not out of my exits. Second, the run never once threatened the daily limit: the worst day was minus 500 dollars, 10% of the limit, and the worst drawdown from a peak was 1,000 dollars. I check those two numbers on every test with the max daily drawdown calculator before I trust a new rule with real size — if a ten-trade test can put 40% of a daily limit at risk on a normal losing run, the rule is too big for the account, whatever its win rate says.
The Mistakes That Cost Me Money
The first mistake was trading the LSMA on M5. I tried a 25-period LSMA on the five-minute chart for one morning, took three trades, and lost two of them to whipsaw before lunch, because at that length the line flips on every minor rotation and the spread alone eats the edge. The LSMA’s low lag is only an advantage when the timeframe is long enough for the trend to mean something; on M5 it is just a faster way to be wrong. H1 is my floor now, and I do not go below it on a funded account.
The second mistake was treating the overshoot as a breakout. After a sharp upside spike the LSMA printed well above price, and I read that gap as strength and bought into it, only to watch price drift back to the line while the line drifted back to price. The gap between a fast line and price is not momentum; it is the maths of a tilted fit unwinding. Now, when the LSMA floats in empty space, I wait for it to rejoin price before I take any signal at all.
The third mistake was the one I already mentioned: taking a CPI-morning long 40 pips above the real market because the news spike had tilted the line. The news filter in my settings table exists because of that trade, and it is non-negotiable — no entries 30 minutes either side of a red-folder release, because the LSMA’s worst hour is the hour after big news. The fourth mistake was subtler: I sized a trade against the daily slope at full 0.5% risk, the H1 line flipped twice, and I lost two half-days of room before lunch. Against the daily slope I trade 0.25% now, and the daily-slope row in the settings table is the rule that enforces it.
Frequently Asked Questions
What is the least squares moving average in simple terms?
It is a moving average that fits a straight line through the last N closes and plots where that line ends right now. Because the line can tilt, its endpoint sits closer to current price than a same-length SMA or EMA, so it turns earlier at trend changes. It is also called the endpoint moving average or the linear regression value.
What LSMA settings work best for prop firm trading?
LSMA 25 on H1 as the trigger and LSMA 50 on H4 as the trend filter is the combination I tested for ten days with 0.5% risk per trade. Shorter lengths such as 9 or 14 flip too often on a funded account, and anything above 50 on H1 turns too slowly to be worth the switch from an EMA.
Is the LSMA better than the EMA?
It is faster, not better in every way. On the same ten closes the LSMA sat 1.4 pips below price while the EMA sat 13.4 pips below, so the LSMA signals earlier. The price of that speed is overshoot after spikes and jumps when old bars leave the window, which the EMA does not have. Use the LSMA when early entries matter more than smoothness, and the EMA when you can afford to be late.
How is the LSMA different from the linear regression channel?
The channel plots bands around a regression line to show where price is stretched; the LSMA is the line itself, used as a trigger through its slope and its position relative to price. Same maths at the core, different job on the chart. I cover the channel separately because its rules, bands and stretch readings deserve their own guide.
Can I use the LSMA on forex pairs with no real volume?
Yes, and that is one of its strengths for funded accounts. The LSMA is calculated from closes only, so it needs no volume feed and the tick-volume problem on spot forex CFDs never affects it. Volume-based tools need a real volume feed to be trustworthy; the LSMA does not.
What is the biggest mistake traders make with the LSMA?
Trading its overshoot. After a sharp spike the line can print well away from price, and buying that gap is buying a maths artefact, not momentum. Wait for the line to rejoin price, check the higher-timeframe slope, and never take an LSMA signal within 30 minutes of a red-folder news release.
Risk disclaimer: This article is educational only and not financial advice. Prop firm accounts are simulated evaluations with strict loss limits; never risk money you cannot afford to lose, and always check your firm’s current rules before trading.




