MACD, RSI and IBS — How SatoshiMedia Predicts Polymarket 15-Minute Contracts

SatoshiMedia's V3 signal engine uses exactly three indicators operating across three different timeframes. Each one answers a different question about the market. All three must align before a signal fires on Polymarket 15-minute binary contracts. This article is a complete technical deep-dive into how each indicator works, why specific parameters were chosen over the textbook defaults, how the indicators combine into a single signal, and where the edge actually comes from.

The core strategy is mean reversion — identifying moments when price has temporarily moved against the dominant trend and is likely to snap back within the 15-minute window. A secondary mode captures early directional momentum at the start of a window when the Polymarket contract has not yet priced in the move. Both modes are built on the same three indicators; the difference is in how they combine them.

Before getting into each indicator, it is worth stating the philosophy clearly: no single technical indicator works reliably on short crypto timeframes. Every indicator produces false signals, lags, or breaks under regime change. The only way to extract signal from noise is confluence — requiring multiple independent indicators to agree before taking action. The mathematics of confluence is discussed later in this article, but the short version is that combining two 60%-accurate independent signals produces a combined accuracy substantially above either alone. This is why SatoshiMedia's approach is explicitly multi-indicator, multi-timeframe, and built around category diversity rather than redundant confirmation.

Why technical indicators work at all on 15-minute crypto

There is a long-running debate over whether technical analysis "works." The honest answer is nuanced: some indicators capture genuine patterns in crowd behavior and market microstructure that persist over time, while others are largely noise that practitioners rationalize post-hoc. The three indicators used by SatoshiMedia — MACD, RSI, and IBS — are in the first category, and each has both empirical support and a clear mechanism for why it should produce edge on short timeframes.

Short-timeframe crypto markets have two key properties that make them receptive to technical analysis. First, participation is dominated by retail traders using similar toolsets, which creates self-fulfilling price patterns — when thousands of traders watch the same RSI levels on the same timeframes, price tends to react at those levels. Second, market microstructure on 15-minute windows is driven by the interaction of algorithmic market makers (who respond to mean reversion) and short-term speculators (who chase trends), producing predictable tension and release patterns that indicators can capture.

The edge is neither large nor guaranteed. A well-calibrated three-indicator system might produce a 65-72% win rate on filtered signals — a meaningful edge over the 50/50 base rate, but not an overwhelming one. Combining that accuracy with disciplined expected value filtering and good bankroll management is what turns a modest analytical edge into sustainable profitability.

Indicator 1: 1-Hour MACD(8,17,9) — Trend direction

The Moving Average Convergence Divergence indicator, developed by Gerald Appel in the late 1970s, answers one question on the 1-hour timeframe: which direction is the market trending right now? It is the trend filter that determines whether SatoshiMedia will even consider an UP signal or a DOWN signal for the current 15-minute window. If MACD says the market is bearish, the scanner will not generate UP signals no matter what RSI and IBS indicate.

The math behind MACD

MACD is built from three exponential moving averages. The MACD line is the difference between a fast EMA and a slow EMA of the closing price: MACD = EMA(fast) − EMA(slow). The signal line is an EMA of the MACD line itself: Signal = EMA(MACD, signal_period). The histogram — what SatoshiMedia actually reads — is the difference between the two: Histogram = MACD − Signal. When the histogram is positive, the MACD line is above the signal line and momentum is bullish. When negative, bearish.

SatoshiMedia uses custom parameters (8, 17, 9) instead of the classic (12, 26, 9). The textbook values were chosen decades ago for daily equity charts, where a 26-day slow EMA is roughly a month of trading — the right horizon for long-term trend filtering on stocks. For 1-hour crypto bars feeding a 15-minute signal window, a 26-bar slow EMA represents 26 hours of price action, which is too slow to detect trend shifts in time for the next 15-minute contract.

The (8, 17, 9) parameterization shortens the fast EMA to 8 bars and the slow to 17 bars, roughly equivalent to 8 and 17 hours of crypto price action. This captures trend shifts within a few hours rather than a full day, which matches the horizon over which 15-minute contracts are meaningfully predictable. The signal line stays at 9 because it controls the smoothing of the histogram, and 9 bars has become a de facto standard that produces stable-but-responsive histogram readings.

MACD strength: the dynamic-threshold driver

Beyond simple direction, the engine calculates MACD strength — the current histogram value relative to its rolling average over the last 20 bars. Strength above 1.5 means the current histogram is significantly larger than recent norms, indicating a powerful trend. Strength below 0.8 means the current histogram is unusually small, indicating a weak or fading trend. This strength value drives the dynamic thresholds applied to the other two indicators.

The logic is this: in a strong trend, even mild pullbacks tend to reverse quickly because the dominant force is pushing one direction. An RSI dip to 40 in a strong uptrend is likely to bounce — the trend does not need extreme conditions to reassert itself. In a weak trend, only deep extremes should be trusted, because the trend does not have enough power to force a quick reversal from marginal oversold conditions. The dynamic threshold matrix reflects this: the stronger the trend, the looser the entry criteria for mean-reversion signals.

This adaptive behavior is what separates a naive three-indicator system from a well-engineered one. A fixed RSI threshold of 30 would produce too few signals in strong trends (missing real opportunities) and too many signals in weak trends (fishing in unfavorable conditions). The dynamic threshold tunes the system to the current market regime automatically.

Indicator 2: 5-Minute RSI(7) — Overbought/oversold detection

The Relative Strength Index, developed by J. Welles Wilder and published in his 1978 book New Concepts in Technical Trading Systems, measures the magnitude of recent price changes to identify overbought and oversold conditions. Mathematically, RSI = 100 − 100/(1 + RS), where RS is the ratio of average gains to average losses over the lookback period. Values range from 0 to 100. Traditional interpretation: below 30 is oversold, above 70 is overbought.

Why 5-minute and why period 7

SatoshiMedia uses RSI(7) on the 5-minute timeframe — two deliberate departures from the classic RSI(14) on daily or hourly bars. Each departure is justified by the specific task of predicting the next 15-minute window.

The 5-minute timeframe is a sweet spot between noise and lag. One-minute RSI triggers constantly on individual candle noise — any single sharp move can send it to extreme values, producing false signals that dissolve within two minutes. Fifteen-minute RSI moves too slowly: by the time it signals oversold, the 15-minute window is nearly over and there is insufficient time for the mean reversion to play out. Five-minute bars capture meaningful price action (roughly $300-$2000 of BTC movement per bar during normal sessions) without being dominated by single-transaction noise, and they produce RSI readings that refresh every 5 minutes — frequently enough to catch dips within a 15-minute window but not so frequently that they trigger on every tick.

The 7-period lookback instead of the textbook 14 also matters. RSI(14) on 5-minute bars spans 70 minutes of price action — longer than the 15-minute prediction window the signal is trying to act on. By the time RSI(14) hits oversold, the relevant dip has been in progress for an hour, and the mean reversion may already have occurred. RSI(7) spans 35 minutes, which is roughly twice the prediction window — long enough for meaningful context, short enough to be responsive to fresh setups.

How RSI feeds the signal

In an uptrend (MACD bullish), RSI below the dynamic threshold indicates an oversold condition — price has dipped against the trend and a bounce is likely. In a downtrend (MACD bearish), RSI above the threshold indicates overbought conditions — price has rallied against the trend and a pullback is expected. This is textbook mean reversion: trading the snap-back to trend after a temporary deviation against it.

The depth of the RSI extreme directly affects signal confidence. An RSI of 10 (deeply oversold in an uptrend) adds up to +10% confidence because historical data shows deeper extremes produce stronger reversions. An RSI of 25 adds +7%. An RSI of 34 (barely crossing the threshold) adds only +1% — it qualifies, but barely. This graduated scaling means the system naturally favors the highest-probability setups rather than treating every threshold crossing equally.

It is worth noting what RSI does not measure: RSI says nothing about the absolute price level, the size of the move, or the volume behind it. It measures relative strength of recent gains versus losses, which is a bounded momentum metric. Two moves of identical dollar magnitude can produce very different RSI readings depending on the prior context. This is a feature, not a bug — it means RSI normalizes across different volatility regimes, giving comparable readings on a quiet day and a volatile day.

RSI divergence: what SatoshiMedia does not use

Traditional technical analysis pays a lot of attention to RSI divergence — situations where price makes a new high but RSI does not, or vice versa. Divergence is a popular concept, but for 15-minute prediction windows it is too slow to be actionable. Divergence patterns typically play out over many bars, which on a 5-minute chart means 30-60 minutes — beyond the horizon of a single 15-minute contract. SatoshiMedia's engine ignores divergence and focuses on simple threshold crossings with depth weighting.

Indicator 3: 15-Minute IBS — Mean reversion confirmation

Internal Bar Strength (IBS) is the least known but potentially most powerful of the three indicators. It was popularized in quantitative trading circles by Larry Connors and appears in academic research on short-term mean reversion in equities. The formula is simple and elegant: IBS = (Close − Low) / (High − Low). It measures where the candle closed relative to its intraday range, producing a value between 0.0 and 1.0.

An IBS of 0.0 means the candle closed at its absolute low — sellers pushed price down and held it there through the close. An IBS of 1.0 means it closed at its absolute high — buyers dominated the close. An IBS of 0.5 means the close was at the midpoint of the range, indicating balanced flow.

Why IBS predicts mean reversion

The empirical regularity that makes IBS valuable is this: on liquid markets, candles that close near their low (IBS < 0.3) tend to be followed by positive returns in the next period, and candles that close near their high (IBS > 0.7) tend to be followed by negative returns. This has been documented across equity indexes, individual stocks, and crypto, with the effect size varying by market. In equities, backtested win rates for IBS-based mean reversion strategies have historically exceeded 65% when combined with a trend filter.

The mechanism is microstructural. When a candle closes near its low, it typically means selling pressure was intense through the closing minutes — often because stop-losses were hit, forced selling occurred, or momentum traders piled in. That pressure tends to exhaust itself at the close, and the next period opens without the same selling force. Buyers re-enter, and price reverts toward the range midpoint. The mirror image applies for high IBS candles.

This is distinct from trend-following, momentum, or breakout strategies — all of which predict that the direction of the last bar will continue. IBS predicts the opposite: that extreme closes are exhaustion, not confirmation. Combining IBS with a trend filter (MACD) resolves the apparent contradiction: the trend filter tells you which direction the market is actually moving, and IBS tells you when a counter-trend move against that direction has exhausted itself.

Previous candle, not current

SatoshiMedia uses the IBS of the previous completed 15-minute candle, never the current partial candle. This is critical for two reasons. First, the previous candle is fully formed and its IBS is a settled fact — using it means the signal criteria are deterministic and reproducible. Second, a partial candle's IBS changes every tick as new high/low/close values are recorded, so any signal based on it would flicker unreliably. The completed previous candle gives a stable input that correctly represents "how did the last period close."

For an UP signal, the previous 15-minute candle must have a low IBS (closed near its low) — indicating that sellers pushed price down, but the broader trend (confirmed by 1-hour MACD) is still bullish. The snap-back from that selling exhaustion is what the bet captures. For a DOWN signal, the reverse: the previous candle closed near its high despite a bearish trend, suggesting buyers are about to exhaust.

Why confluence works: the mathematics

Combining three independent indicators is not arbitrary — it reflects a mathematical property that makes the combined signal substantially more accurate than any individual indicator. Suppose each indicator, taken alone, produces a 60% correct directional call. If the three indicators were perfectly independent, the probability that all three agree and are correct is governed by Bayes' theorem. Without going into the full derivation, the key result is that when multiple independent noisy signals agree, the combined signal's accuracy compounds upward much faster than it would for a single stronger signal.

The critical word is independent. If all three indicators measure the same underlying thing — for instance, if they are all variants of momentum — they will almost always agree, and their agreement provides no extra information beyond one of them alone. This is called signal redundancy, and it is the most common mistake in multi-indicator systems. Traders stack 5 momentum indicators, see them all "confirm" the trend, and think they have strong confluence. In reality, they have one signal repeated five times.

SatoshiMedia's three indicators are explicitly chosen to measure different things. MACD is a trend indicator — it describes the direction and strength of the dominant price movement over a meaningful horizon. RSI is a momentum indicator — it describes whether recent gains or losses have been disproportionately large relative to their typical magnitude. IBS is a microstructure indicator — it describes where the most recent candle closed within its own range, capturing exhaustion patterns that neither trend nor momentum can see. The three categories (trend, momentum, microstructure) capture largely independent information, which is what makes confluence meaningful rather than redundant.

A signal that has MACD trend direction, RSI at a meaningful extreme, and IBS confirming exhaustion is a signal where three different views of the market are all pointing to the same conclusion. That is the kind of agreement that moves the probability estimate materially above 50/50.

Multi-timeframe analysis: why 1H + 5m + 15m

The choice of three different timeframes (1-hour MACD, 5-minute RSI, 15-minute IBS) is not arbitrary either. Each timeframe captures a different horizon of market activity, and their alignment across timeframes is stronger evidence than alignment on a single timeframe.

The 1-hour MACD captures the dominant regime. This is the horizon over which trend-following algorithms operate, where institutional flows leave visible footprints, and where macroeconomic news has had time to affect price. Trading against the 1-hour MACD direction is trading against the strongest signal available about which way the market is leaning. Using it as a filter eliminates a large class of bad setups — trades that look good on a 5-minute chart but run straight into the 1-hour trend.

The 5-minute RSI captures tactical timing within the regime. Once MACD has set the direction, RSI tells you when prices have stretched far enough against that direction to expect a snap-back. The 5-minute timeframe is responsive enough to catch dips that materialize within a single 15-minute window, without being so responsive that it triggers on every minor wiggle.

The 15-minute IBS captures microstructure evidence that the recent dip is actually exhaustion rather than a new directional move. A low IBS after an RSI dip means the sellers who caused the dip have been pushed to their limit — they took price to the low and could not hold it higher through the close. That is very different from a dip followed by a balanced close, which might indicate sellers are still in control.

Together, the three timeframes form a coherent picture: 1-hour regime, 5-minute stretch, 15-minute exhaustion. A signal where all three align is telling the same story from three different vantage points — the market is trending in direction X, has temporarily moved against X, and the counter-move has exhausted. That is when edge is maximum.

Two signal modes: Mean Reversion and Early Momentum

The three indicators power the primary signal mode — Mean Reversion — as described above. All three must align: MACD sets the trend, RSI confirms the dip, IBS confirms the exhaustion. This mode requires at least 7 minutes remaining in the window to allow time for the reversion to play out. Signals firing with less than 7 minutes left are suppressed because the mean-reversion move often takes 3-8 minutes to complete, and a 5-minute window is statistically unreliable.

The secondary mode — Early Momentum — activates only during the first 5 minutes of a new window, when the Polymarket entry price is still below $0.55 (meaning the market is genuinely uncertain about direction). Instead of RSI and IBS extremes, Early Momentum requires three consecutive 5-minute candles moving in the same direction as the MACD trend, confirmed by above-average volume. This captures situations where strong directional momentum has not yet been priced into the Polymarket contract — the book is still near 50/50 even though the 5-minute chart clearly shows directional flow.

Early Momentum signals receive a lower base confidence than Mean Reversion signals — approximately 5 percentage points less. This reflects the empirical reality that trend continuation is less reliable than mean reversion from exhaustion extremes, plus the fact that 5-minute consecutive candles can be misleading during news-driven spikes that revert almost immediately. The Early Momentum mode is a supplementary source of signals during the periods when Mean Reversion is waiting for indicator alignment, not a replacement for the primary mode.

The interaction between the modes is designed to be additive, not redundant. Early Momentum can only fire in the first 5 minutes of a window. Mean Reversion can only fire between minutes 5 and 12 (requiring 7 minutes for resolution). The last 3 minutes of any window produce no new signals, as edge erosion and execution risk make late-window trades unfavorable. In practice, this means any given 15-minute window produces at most one fresh signal per asset — reducing the temptation to overtrade on partial alignments.

Secondary inputs: volume, session quality, orderbook imbalance

The three primary indicators are the core of the signal engine, but several secondary inputs refine the confidence estimate and filter out marginal conditions.

Volume. The engine computes a volume multiplier as the current 5-minute volume divided by the 20-bar average. A volume multiplier above 1.3 adds a small confidence bonus (+3%) because price moves on elevated volume are more likely to reflect genuine conviction rather than thin-book noise. Very low volume (below 0.6×) applies a confidence penalty because indicator readings in low-volume conditions are less statistically reliable.

Session quality. Crypto trades 24/7, but the quality of technical signals varies by session. US market hours (13:00-21:00 UTC) tend to produce the highest-quality signals because equity market open influences crypto flow, institutional algorithms are most active, and order book depth is maximum. Late-night hours (02:00-05:00 UTC) tend to produce lower-quality signals because volume is thin, spreads are wider on Polymarket, and indicator readings are more noise-sensitive. The session quality multiplier reduces confidence for off-peak signals by 3-7 percentage points depending on the specific hour. This is covered in more detail in best times to trade crypto prediction markets.

Polymarket orderbook imbalance. The engine also reads Polymarket's own order book for the 15-minute contract. Heavy bid-side imbalance (more than 2× more bids than asks across the top 5 levels) is a soft confirmation that informed participants are positioned in the same direction as the signal. Heavy ask-side imbalance is a warning sign — even if the technical indicators all align UP, strong selling pressure in the contract itself suggests other informed participants disagree with the directional call. Orderbook confirmation adds up to +2% confidence; orderbook contradiction subtracts up to -3%.

These secondary inputs never create signals on their own. They exist to modulate the confidence of signals that have already passed the three-indicator alignment test. A signal without volume, with poor session quality, and with contradictory orderbook would still fire if the primary indicators align, but its confidence would be at the low end of the allowed range — and in most cases its EV would fall below the action threshold.

Confidence calculation and expected value

Every signal includes a confidence score (estimated win probability) and an expected value. The confidence starts from a base value — initially 0.63 (synthetic prior), which adapts over time as the system tracks its own resolved signals. After 20 resolved signals, the base shifts to a blend: 70% recent realized win rate plus 30% prior. After 100 resolved signals, the prior has essentially faded and the base is dominated by actual performance. This is a simple self-correcting mechanism that prevents the system from indefinitely overestimating its own accuracy.

On top of the adaptive base, bonuses and penalties adjust the per-signal confidence. RSI depth adds +1% to +10% depending on how far below (or above) threshold it is. IBS extremity adds up to +5% for readings very close to 0.0 or 1.0. MACD acceleration (histogram larger than the previous bar's) adds +3%. Volume spikes add up to +3%. Polymarket orderbook confirmation adds +2%. Penalties include session quality (-3% to -7% for off-peak hours), BTC counter-signals for altcoins (-3% when BTC and the altcoin point in opposite directions), and wide Polymarket spreads (-2% when the bid/ask spread exceeds $0.04).

The final confidence is clamped between 52% and 82% — preventing unrealistic extremes. The floor at 52% exists because any signal surviving the three-indicator alignment has at least some edge over random, and pushing confidence below 52% would imply the signal filter failed. The ceiling at 82% exists because no combination of technical indicators can push true probability meaningfully above this level on a 15-minute crypto contract — beyond 82%, the confidence estimate is almost certainly overfitting rather than real edge.

Expected value is then calculated using the confidence and the current Polymarket contract price: EV = (confidence × (1 − price) × (1 − 0.02)) − ((1 − confidence) × price). A signal is only emitted when EV exceeds +$0.04 per dollar and the edge over breakeven exceeds 2 percentage points. See the EV guide for a complete discussion of how EV translates into long-run profitability.

Resolution: Polymarket as the only source of truth

Every signal's WIN/LOSS outcome is determined exclusively by querying Polymarket's own API after the 15-minute window closes. The scanner waits 3 minutes for Polymarket to resolve the market via its Chainlink oracle, then checks whether the YES or NO side won. No Binance price data, no manual adjudication, no interpretation — the result comes from the same source where the bet would have been placed.

This matters because Polymarket's resolution can occasionally differ from what Binance shows due to oracle timing, data-source aggregation differences, or rounding conventions. Polymarket's oracle typically uses a composite of major exchanges, which may show a slightly different close than any single exchange. By resolving from Polymarket directly, the tracked win rate reflects exactly what a user who followed the signal would have experienced. The tracker is not "correct on paper" — it is correct against the same authority that settles real bets.

This resolution discipline also prevents a subtle form of backtesting bias. Systems that use a different price source for resolution than for execution can appear to have higher win rates than they would deliver in live trading, because execution and resolution can drift apart by fractions of a percent. SatoshiMedia eliminates this gap entirely by using Polymarket's settlement as both the trading venue and the resolution source.

Limitations and when the system fails

No signal system is universally applicable. There are specific conditions under which the three-indicator approach works poorly, and understanding them is essential for deciding when to trust the signals and when to stand aside.

Regime breaks. The system is calibrated on historical data from specific volatility regimes. When crypto markets transition sharply — for example, when a prolonged low-volatility consolidation breaks into a volatile trend — the dynamic thresholds may be mis-calibrated for a few days until the rolling MACD strength average catches up. During transition periods, signal quality temporarily degrades.

News-driven moves. Technical indicators describe historical price action. When a headline hits — a Fed decision, a major exchange hack, an ETF announcement — price moves immediately on information the indicators did not have. A signal that fires 30 seconds before a news event is acting on stale information. The system cannot filter news risk; it relies on the user's judgment to avoid trading around known scheduled events.

Extremely thin markets. On weekends, holidays, and Asian late-night hours, Polymarket order books for some assets (especially BNB) become very thin. In these conditions, the Polymarket contract price can move significantly on a single order, and the indicator-derived edge is often smaller than the slippage cost. Session-quality filters reduce signal emission during thin-market periods, but users should still be aware that thin conditions degrade execution.

Mean reversion vs breakout regimes. The core strategy is mean reversion — betting that counter-trend moves will reverse. This works well in trending-but-bounded markets (the normal crypto condition) and poorly in strong breakout conditions where dips keep getting deeper because a new leg is forming. The MACD trend filter partially addresses this by requiring the broader trend direction to support each trade, but it cannot fully eliminate the risk of trading into a breakout that does not revert.

Cross-asset correlation collapses. Most of the time, BTC, ETH, SOL, and BNB move in the same direction. Occasionally — for example, during asset-specific news or exchange-listing events — one asset decouples sharply. The BTC counter-signal filter catches most cases, but genuine decoupling can produce signals on one asset that fail because the broader crypto move pulls the wrong way.

Using the indicators in practice

You can see all three indicators in action on the SatoshiMedia dashboard, where each coin card shows RSI 5m, MACD 1H, IBS 15m, session quality, and Polymarket orderbook imbalance in real time. The signal layers display shows how many of the three required indicators are currently aligned for each asset. When all three light up and the expected value passes the threshold, a signal fires, a Telegram alert goes out to the private Telegram channel, and the clock starts ticking on the next 15-minute contract.

For deeper context on the specific assets, the Bitcoin, Ethereum, Solana, and BNB prediction pages show the live indicator readings alongside the 7-day rolling win rate for that asset. Comparing realized performance to the predicted confidence is the fastest way to calibrate your trust in the system — if the realized rate matches the predicted rate over 100+ resolved signals, the system is well-calibrated for that asset. If there is a persistent gap, raise the EV threshold before acting.

Users who prefer to verify signals manually can open the embedded TradingView chart on each asset page, where the same Binance price feed the scanner uses is available with professional charting tools. Adding MACD(8,17,9) on the 1-hour chart and RSI(7) on the 5-minute chart reproduces the exact indicator readings the scanner sees, making the signals fully transparent.

Putting it together

The three-indicator approach — 1-hour MACD for trend, 5-minute RSI for mean-reversion timing, 15-minute IBS for microstructure confirmation — is the analytical core of SatoshiMedia's Polymarket 15-minute prediction engine. Each indicator answers a different question, each operates on a different timeframe, and their confluence produces a signal substantially stronger than any component alone. Dynamic thresholds adapt the system to changing trend strength, secondary inputs refine the confidence estimate, and Polymarket-based resolution guarantees that tracked performance reflects real execution reality.

The edge is real but modest — roughly 65-72% win rate on filtered signals in typical conditions, occasionally higher during strong trends, occasionally lower during regime transitions. Converting that edge into sustainable profit requires disciplined EV filtering, appropriate position sizing, and acceptance of the variance that accompanies any probabilistic strategy. The indicators generate the setups; the trader's discipline determines the outcome.