TrendAviv Why add Tower Semiconductor to coverage?
Because it broadens semiconductor exposure on the site beyond one business model.
Equities
[Horizon: T+1 Close]
SELL
Tower Semiconductor TSEM
Forecasting and modeling
Latest reference level
119.44
Active-horizon target
118.82
-0.52%
Raw interval (bands)
S1 108.40
R1 130.48
R1 130.48
Volatility (σ)
66.8%
Assessment (Yesterday for Today): 119.98
Model quality (out-of-sample)
84.9%
Out-of-sample sample: 200 sessions
Why this asset matters
Tower Semiconductor adds a different semiconductor angle to the Israeli technology universe: foundry and manufacturing exposure rather than process-control tooling. It helps broaden the site’s view on how the global chip cycle is feeding into local listed equities.
What usually moves this asset
Tower is commonly influenced by foundry demand, customer-cycle expectations, global semiconductor sentiment, Nasdaq technology tone and local risk appetite for growth names.
How to read this page
Use the public signal here as a short-horizon read on a foundry-linked Israeli technology stock. It is especially useful in comparison with Nova and the broader local tech basket.
Methodology and dashboard reading (Model Card)
Model architecture:
Tree-ensemble architecture (Random Forest / gradient-boosted logic) built on price structure, volatility regime, macro-relative features, cross-asset context and sentiment inputs. For non-FX assets, the model is not centered on rate differentials alone.
Tree-ensemble architecture (Random Forest / gradient-boosted logic) built on price structure, volatility regime, macro-relative features, cross-asset context and sentiment inputs. For non-FX assets, the model is not centered on rate differentials alone.
Horizons (Targets):
T+1: Direct forecast for the currently displayed horizon close.
T+30: Direct horizon forecasts for each trading day up to T+30, displayed as a path.
1M-1Y: asset_model_targets_1m_1y_nonfx
T+1: Direct forecast for the currently displayed horizon close.
T+30: Direct horizon forecasts for each trading day up to T+30, displayed as a path.
1M-1Y: asset_model_targets_1m_1y_nonfx
Validation:
Walk-forward testing on expanding history without future leakage. Retraining is performed inside each evaluation step.
Walk-forward testing on expanding history without future leakage. Retraining is performed inside each evaluation step.
- Entry: At spot price (snapshot before model calculation). Exit: At active-horizon close.
- Divergence: If the technical trend is falling but the signal is BUY, the model has detected a local oversold condition and is trading a rebound (mean reversion).
- Execution risk: The signal is not investment advice and does not include sizing. The model does not account for broker spreads, slippage, or carry costs (swap/rollover for FX).
- News risk: At extreme risk sentiment (GDELT), raw bands are not calibrated for outliers. The probability of fat tails increases.
1. Current market regime (Technical & Macro View)
Heuristic regime gauges
1W
Strong decline
1M
Strong decline
6M
Neutral
Heuristic regime display for page context, not a price-return backfill.
Heuristic regime indicators
Regime oscillator Displayed as a heuristic regime oscillator for page context. In the current version, this is not a full classical RSI calculation from the complete indicator pipeline.
10.0 (Oversold)
Trend-bias proxyBearish
Macro / carry biasBelow SMA 50/200
Volatility (σ)66.8%
Display heuristics for current regime reading. Not a full classical technical-indicator calculation.
Volatility (σ)66.8%
Macro factors (Percentile Score)
Values are normalized from 0 to 100
Sentiment (GDELT NLP)
Tension indexN/A
StatusData unavailable
Source: GDELT Project API
Coverage: ~400 media / 24h
Updated: 25.10.2026 02:30 UTC
News-tone classification, not a factual confirmation of the event.
Coverage: ~400 media / 24h
Updated: 25.10.2026 02:30 UTC
News-tone classification, not a factual confirmation of the event.
2. Deep quantitative analytics
Historical seasonality
This panel is temporarily hidden while historical monthly returns are being rebuilt from the same source-aligned series used by the model.
A real seasonality heatmap will return after monthly returns are recalculated from the same source-aligned history used by the model.
Medium-term continuation estimate
For indices, equities, bonds and commodities, the 1M-1Y block is shown as a continuation of the 30-trading-day ML anchor, not as an FX-style CIP fair-value model.
Medium-term scenario: 102.25
Inputs: 30-trading-day ML anchor + medium-term continuation profile
30D anchor: 102.25
Extension profile: Anchor continuation with volatility widening
Current level: 119.4430D anchor: 102.25
Extension profile: Anchor continuation with volatility widening
Medium-term scenario: 102.25
Mean-revertContinue
Gap versus 30D anchor: 0.00%
Model verdict: Balanced continuation
3. ML model assessment (Predictive Analytics)
Historical validation: model forecast vs reality
Bands are nominal (raw). Current empirical out-of-sample coverage: 89.5%. The model is still being calibrated.
For historical validation, this chart uses the same source as the model input: TASE / Yahoo Finance. This removes cross-source shift.
Return-space check: predicted vs realized active-horizon move
This view removes the false level-lag effect and shows the predicted change from the previous close.
ML model summary (tree ensemble)
Analyzing macroeconomic data, the ML model detects a confident bearish (downtrend) regime on the medium-term horizon. The news-sentiment layer is unavailable in this refresh and is excluded from the narrative interpretation. RSI (10.0) confirms the local structure.
Active-horizon signal: SELL
Return to target: -0.52%
Return to target: -0.52%
Attention (Risk calibration):
Raw interval. Empirical OOS coverage: 89.5%. Calibration is still required (fat tails).
Raw interval. Empirical OOS coverage: 89.5%. Calibration is still required (fat tails).
Directional model accuracy estimate 56.5%
Empirical coverage 89.5%
MAE edge Baseline (naive forecast) = tomorrow's price equals today's price. Edge = (MAE_naive - MAE_model) / MAE_naive.
MAE Model: 2.3381
MAE Naive: 2.2517+1.0% Edge
MAE Model: 2.3381
MAE Naive: 2.2517+1.0% Edge
Out-of-sample sample Based on 200 recent out-of-sample sessions shown on this page. Small samples can materially overstate or understate stability.200 sessions
4. Probabilities and trading signals
Close probability over a 30-trading-day horizon Mode: Normal (Calm)
Estimate of the probability that price closes above or below target levels exactly after 30 trading days (T+30 sessions).
Calculated via Gaussian CDF N(μ, σ). Does not account for skewness or fat tails during geopolitical FX shocks.
Close ABOVE level (Upside)
Level R1 (130.48)11%
Level R2 (132.27)9%
Close BELOW level (Downside)
Level S1 (108.40)61%
Level S2 (106.61)58%
Trading interpretation of the signal:
- Instrument: TSEM
- Entry: At current session close (04:30).
- Exit: At active-horizon close (26.10).
- Model: SELL (Expected return: -0.52%)
Signals are produced exclusively by the ML core and are not investment advice.
Attention: The daily forecast (left) is displayed as direct horizon estimates for each trading day up to T+30. The long-horizon block (right) is a structural scenario layer, so the trajectories can diverge.
Short-horizon ML expectations (daily)
| Date | ML target | Raw interval (bands) | Signal |
|---|---|---|---|
| 26.10.2026 (Mon) | 118.82 | [110.96 - 127.25] | SELL |
| 27.10.2026 (Tue) | 118.21 | [107.30 - 130.23] | SELL |
| 28.10.2026 (Wed) | 117.60 | [104.44 - 132.41] | SELL |
| 29.10.2026 (Thu) | 116.99 | [102.02 - 134.16] | SELL |
| 30.10.2026 (Fri) | 116.39 | [99.86 - 135.65] | SELL |
| 02.11.2026 (Mon) | 115.79 | [97.90 - 136.93] | SELL |
| 03.11.2026 (Tue) | 115.19 | [96.10 - 138.07] | SELL |
| 04.11.2026 (Wed) | 114.59 | [94.41 - 139.08] | SELL |
| 05.11.2026 (Thu) | 114.00 | [92.83 - 140.00] | SELL |
| 06.11.2026 (Fri) | 113.41 | [91.33 - 140.83] | SELL |
| 09.11.2026 (Mon) | 112.82 | [89.90 - 141.60] | SELL |
| 10.11.2026 (Tue) | 112.24 | [88.54 - 142.29] | SELL |
| 11.11.2026 (Wed) | 111.66 | [87.23 - 142.94] | SELL |
| 12.11.2026 (Thu) | 111.09 | [85.97 - 143.53] | SELL |
| 13.11.2026 (Fri) | 110.51 | [84.76 - 144.08] | SELL |
| 16.11.2026 (Mon) | 109.94 | [83.60 - 144.59] | SELL |
| 17.11.2026 (Tue) | 109.37 | [82.47 - 145.06] | SELL |
| 18.11.2026 (Wed) | 108.81 | [81.37 - 145.49] | SELL |
| 19.11.2026 (Thu) | 108.24 | [80.31 - 145.90] | SELL |
| 20.11.2026 (Fri) | 107.69 | [79.28 - 146.28] | SELL |
| 23.11.2026 (Mon) | 107.13 | [78.27 - 146.63] | SELL |
| 24.11.2026 (Tue) | 106.58 | [77.29 - 146.95] | SELL |
| 25.11.2026 (Wed) | 106.03 | [76.34 - 147.25] | SELL |
| 26.11.2026 (Thu) | 105.48 | [75.41 - 147.53] | SELL |
| 27.11.2026 (Fri) | 104.93 | [74.51 - 147.78] | SELL |
| 30.11.2026 (Mon) | 104.39 | [73.62 - 148.02] | SELL |
| 01.12.2026 (Tue) | 103.85 | [72.75 - 148.24] | SELL |
| 02.12.2026 (Wed) | 103.31 | [71.91 - 148.44] | SELL |
| 03.12.2026 (Thu) | 102.78 | [71.08 - 148.62] | SELL |
Long-horizon structural expectations
| Horizon end | ML target | Delta | Raw interval (bands) | Signal |
|---|---|---|---|---|
| 25.01.2027 (1 quarter) | 100.25 | -16.06% | [75.15 - 131.07] | SELL |
| 25.04.2027 (6 months) | 97.33 | -18.51% | [61.14 - 147.33] | SELL |
| 25.10.2027 (1 year) | 91.59 | -23.32% | [44.87 - 167.05] | SELL |
Long horizons are shown as a structural scenario layer built from spot, rates, volatility and macro drift. Read them separately from the short-horizon ML forecast.
5. Research sandbox (Illustrative sensitivity)
Stress-test calculator (What-If Analysis)
Illustrates linear sensitivity to shocks (based on heuristic weights). Under stress, the risk corridor widens (volatility).
Short-horizon ML forecast (up to 30 trading days)
Long-horizon structural scenario (up to 1 year)
6. Interactive chart (TradingView)
Attention: the base model calculation uses TASE / Yahoo Finance. The chart below shows the TradingView feed (TASE:TSEM), so small quote differences are possible.
Related market pages
FAQ
How should I read the active-horizon signal here?
As a short-term directional lean into the currently displayed horizon close, always together with volatility and sector context.
Why compare it with Nova?
Because both are tied to the chip cycle but through different business exposures.
Why can global tech sentiment matter so much?
Because semiconductor names are tightly linked to global capex and demand expectations.