Telegaon is a popular website for its price predictions of stocks. Many world-renowned financial institutions have mentioned the work of Telegaon. Before making price predictions of a stock, we use this methodology.
Overview
Our stock price predictions are generated through a multi-layered analytical framework that combines quantitative technical analysis, macroeconomic context, and a machine learning-assisted forecasting model. No single input drives the output — predictions are the result of cross-validating signals across multiple independent methodologies, with the final range reflecting the distribution of probable outcomes rather than a single point estimate.
All predictions are presented as a minimum–average–maximum range to reflect genuine uncertainty. We treat price prediction as a probabilistic exercise, not a deterministic one.
Step 1: Historical Price Analysis
The foundation of every prediction is a structured review of the stock’s historical price behaviour over three time horizons: short-term (90 days), medium-term (1–3 years), and long-term (IPO or 10-year maximum).
Within each horizon we calculate:
Simple Moving Averages (SMA) at the 50-day and 200-day intervals. The relationship between these two averages — specifically whether the 50-day has crossed above or below the 200-day — is used to establish baseline trend direction. A sustained “golden cross” (50-day above 200-day) contributes a bullish weight to the model. A “death cross” contributes a bearish weight.
Exponential Moving Averages (EMA) at the 12-day and 26-day intervals feed into the MACD calculation (see below). EMAs are weighted more heavily than SMAs in our model because they respond faster to recent price changes, which is particularly relevant for high-volatility growth stocks.
Annualised volatility is calculated using the standard deviation of daily log returns over the trailing 252 trading days. This figure directly influences the width of the minimum–maximum prediction band: a stock with 60% annualised volatility will carry wider prediction bands than one with 25%, reflecting genuine uncertainty rather than false precision.
Step 2: Technical Indicator Suite
Six core technical indicators are calculated and assigned a directional signal (bullish / neutral / bearish) before being combined into a composite score:
RSI (Relative Strength Index, 14-period): Readings above 70 signal overbought conditions and contribute a bearish weight. Readings below 30 signal oversold conditions and contribute a bullish weight. The RSI is particularly useful for identifying potential near-term reversals but is treated with lower weight in long-term predictions.
MACD (Moving Average Convergence Divergence): The MACD line (12-day EMA minus 26-day EMA) is compared against a 9-day signal line. A positive MACD histogram that is expanding contributes a bullish signal; a narrowing or negative histogram contributes a bearish signal.
Bollinger Bands (20-period, 2 standard deviations): Price proximity to the upper or lower band informs mean-reversion probability. When a stock trades near or beyond the upper band with declining volume, the model applies a mild bearish correction to near-term targets. A touch of the lower band with increasing volume is treated as a bullish setup.
Volume Trend: Raw price movement without volume confirmation is treated as a weaker signal. We calculate a 20-day volume moving average and compare it against the volume on significant price move days. High-volume breakouts receive stronger bullish weight than low-volume ones.
Support and Resistance Levels: Key historical price levels — defined as price zones where the stock reversed direction at least twice — are identified manually and used to set realistic ceiling and floor estimates within each prediction year. These levels anchor the minimum and maximum figures in the summary table.
52-Week High/Low Proximity: A stock trading within 10% of its 52-week high in an otherwise bullish macro environment receives a modest upward bias. A stock trading within 10% of its 52-week low with deteriorating fundamentals receives a downward bias.
The six signals are combined into a composite score on a scale of -6 (all bearish) to +6 (all bullish). This score shifts the base trajectory estimate up or down before fundamental overlays are applied.
Step 3: Fundamental Overlay
Technical indicators describe price behaviour. Fundamentals describe whether that behaviour is likely to be sustainable. Our model incorporates the following fundamental inputs:
Revenue trajectory: Year-over-year revenue growth rate and the direction of that trend (accelerating vs. decelerating). A company growing revenue at 40% with an accelerating trend receives a positive fundamental multiplier. A company with declining revenue growth is penalised.
Profitability and margin direction: We assess operating margin, EBITDA margin (where applicable), and the trajectory of net losses for pre-profitability companies. For companies that remain unprofitable, we note the expected path to profitability and weigh it against the risk of cash runway depletion.
Analyst consensus: We aggregate available Wall Street analyst price targets from published sources (Yahoo Finance, StockAnalysis, MarketBeat, WallStreetZen) to calculate a consensus 12-month target and the distribution of ratings (Strong Buy / Buy / Hold / Underperform / Sell). The consensus target serves as a reality check against our own model — significant divergence in either direction is flagged and investigated.
Balance sheet health: Debt-to-equity ratio, current ratio, and available cash are reviewed to assess whether the company can sustain its operations and growth plans. Companies with thin cash runways relative to their burn rate receive a risk discount applied to longer-term predictions.
Sector and macro tailwinds/headwinds: We assess the broader industry environment — regulatory changes, interest rate sensitivity, commodity input costs, and competitive dynamics. These are incorporated as qualitative multipliers rather than precise numerical inputs.
Step 4: Machine Learning Price Range Estimation
Historical price data (adjusted for splits and dividends), the technical composite score, and the fundamental overlay inputs are fed into a gradient-boosted regression model trained on comparable assets in the same sector. The model outputs a probability distribution of price outcomes for each prediction year.
From this distribution we extract:
- The minimum (10th percentile outcome under bear-case conditions)
- The average (median predicted price)
- The maximum (90th percentile outcome under bull-case conditions)
We intentionally use the 10th and 90th percentiles rather than the absolute tail extremes because the tails reflect low-probability black-swan scenarios that are not useful for investment planning purposes. The range is wide enough to be honest about uncertainty, but not so wide as to be meaningless.
Step 5: Long-Term Extrapolation (2035, 2040, 2050)
Beyond a five-year horizon, technical analysis becomes largely irrelevant. Long-term predictions are driven primarily by:
Compound growth rate scenarios: We model three growth trajectories — a bear case anchored to the S&P 500’s historical CAGR (~9.25%), a base case anchored to the sector’s historical median, and a bull case reflecting company-specific growth potential if its most optimistic roadmap is executed.
Business model evolution: For companies with significant transformation narratives (a hardware maker entering software licensing, an EV company entering autonomous vehicles), we apply a qualitative forward multiple adjustment to account for the potential margin expansion of new revenue streams.
Structural market assumptions: Long-term predictions assume general economic continuity. They do not attempt to model black-swan economic events, world conflicts, or regulatory interventions that have no current precedent.
All predictions beyond five years are clearly labelled as highly speculative estimates. The confidence interval widens substantially with each additional year, and this is reflected in the width of the minimum–maximum band.
Limitations and Important Disclosures
No predictive model can account for all variables that influence stock prices. Past price behaviour is not a reliable indicator of future results. Our predictions are speculative estimates, not financial advice. Users should treat these figures as one input among many in their own research process, not as a recommendation to buy, hold, or sell any security. We update predictions periodically as new financial data, analyst revisions, and material corporate events become available.
Our Data Sources
- TradingView — Charts and technical analysis
- Nasdaq — Stock market data and indices
- NYSE — New York Stock Exchange data
- Bloomberg — Financial news and market data
- MarketWatch — Real-time market quotes and news
