Stops good charts from becoming bad entries
Signal can surface a beautiful setup, but Valuation prevents Harbor from paying any price just because the technical picture looks clean.
Agent 04 · Valuation
Anchor the price.
Quantify the margin.
The Valuation agent is Harbor’s price-discipline layer. It takes the names that survived Macro, Sentiment, and Signal, then asks whether the current price is actually justified. Its importance to the universe is simple: Harbor should not confuse a good setup with a good price.
Pricing backbone
Pricing context
Harbor uses valuation to turn raw conviction into disciplined price context before capital sizing begins.
Candidates entering valuation
4
From the live Signal shortlist
Simulation depth
Research stack
SIZING 5PCT
Last research refresh
Mar 16, 2026
Staged
Chain position
4 of 6
After Signal · Before Risk
Decision Chain Position
Where Valuation sits.
Valuation is the chain’s fundamental anchor. It does not scan the whole universe. It challenges the names that already earned deeper attention and hands Risk a margin band instead of a vibe.
Creates a sizing anchor for Risk
Risk needs more than conviction. It needs a price frame, a cushion, and a margin band so capital can be sized with discipline instead of optimism.
Raises the quality of the whole universe
The better Valuation filters the shortlist, the less downstream governance has to waste time challenging names that are already too stretched.
Valuation Methods
Three lenses. One anchor.
Valuation should not trust one methodology in isolation. Harbor blends multiple lenses so a single optimistic model cannot dominate the conclusion.
40%
DCF · Discounted Cash Flow
Projects free cash flow across an explicit forecast window, then discounts it through a regime-aware cost of capital so the anchor changes when the environment changes.
Revenue growthMarginsCapexWACCTerminal g
35%
Comps · Relative Multiples
Benchmarks a candidate against the right peer set so Harbor can tell whether a premium is justified, ordinary, or already stretched.
EV/RevEV/EBITDAP/EP/FCFPeer set
25%
Monte Carlo · Simulation
Runs a probabilistic range instead of one deterministic guess, so Valuation can hand Risk a band of outcomes rather than a single heroic number.
Scenario pathsForward horizonP10 / P50 / P90SIZING 5PCT
Blended Output
How the target is built.
This is the kind of blended output Harbor would publish after a candidate earns deeper challenge. It is intentionally transparent about contribution instead of hiding the anchor behind one number.
| Method | Weight | Fair Value | Contribution | vs. Current |
|---|---|---|---|---|
| DCF (Base case) | 40% | $212 | $84.80 | +36.8% |
| Comps (Median) | 35% | $198 | $69.30 | +27.7% |
| Monte Carlo (P50) | 25% | $205 | $51.25 | +32.3% |
| Blended Fair Value | 100% | $205 | $205.35 | +32.5% |
Reference frame · candidate context inspired by AXTI
Margin of Safety
Calibrated conviction.
Valuation does not just say supported or not. It maps the entry into a margin band that later changes how much capital Risk is willing to commit.
Overvalued
Thin
Adequate
Strong
Deep
▼
< 0%0 – 10%10 – 25%25 – 40%> 40%
Live research context
Primary research profileSIZING 5PCT
Median 1Y P50 surface—
Average simulated drawdown—
Paths profitable—
Evaluation Pipeline
From shortlist to price discipline.
Valuation is where Harbor slows down. This is the layer that asks whether a technically attractive idea deserves capital at the current price.
Step 01
Receive
Read the shortlist from Signal and inherit macro posture so valuation assumptions stay aware of the environment they are being applied in.
Step 02
Model
Build the valuation frame: baseline operating assumptions, peer context, and scenario distribution for each candidate.
Step 03
Challenge
Compare price to intrinsic value, relative multiples, and distribution bands instead of accepting one lens at face value.
Step 04
Anchor
Blend the outputs into one anchor and classify the entry into a margin band the rest of Harbor can actually use.
Step 05
Emit
Publish valuation posture, target range, and margin of safety downstream so Risk can size capital against real price discipline.
Data Sources
What feeds the models.
Valuation is the deepest data consumer in the chain. It needs statements, peer context, and simulation scaffolding before it can responsibly challenge price.
| Source | Data | Frequency | Provider |
|---|---|---|---|
| Financial statements | Income, balance sheet, and cash flow history | Quarterly | FMP / SEC |
| Consensus estimates | Revenue, EPS, and margin expectations | Daily | FMP |
| Peer multiples | Relative pricing by sector and growth profile | Daily | FMP / Polygon |
| Monte Carlo policy surface | Distribution guardrails and risk ladder context | Model | Harbor research stack |
| Macro posture | Rates and regime context for discount discipline | Live | MacroContext |
| Signal shortlist | Only names that earned deeper challenge | Live | SignalContext |
Output Schema
What Valuation emits.
Valuation should publish a reusable object, not a hidden analyst opinion. Risk and Execution both need the anchor, the band, and the reasoning.
ValuationContextharbor_context.py
{
"ticker": str,
"current_price": float,
"blended_fair_value": float,
"margin_of_safety": float // pct,
"conviction_tier": str // OVERVALUED | THIN | ADEQUATE | STRONG | DEEP,
"dcf_value": float,
"comps_value": float,
"mc_p10": float,
"mc_p50": float,
"mc_p90": float,
"assumption_notes": list[str],
"published_at": datetime
}
Configure & Interact
Calibrate the anchor.
Valuation should be tunable without becoming arbitrary. Harbor keeps the inputs editable, the method stack visible, and the override trail auditable.
01
Adjust method weights
Shift the blend between DCF, relative comps, and Monte Carlo so Harbor can emphasize the right lens for the type of company being challenged.
02
Override assumptions
Tighten or relax growth, margin, and discount inputs for names that need operator judgment, while preserving an auditable trail of overrides.
03
Define peer groups
Control which companies should count as fair comparison so Harbor is not anchoring a candidate against the wrong market neighborhood.
04
Query the anchor
Ask why Harbor sees a candidate as supported, stretched, or rich and get the answer in plain language tied back to the valuation frame.
Try it out
Test the pricing discipline.
Explore the research surface behind Harbor’s anchor, run a live name through Insight, or continue into Risk to see how price discipline becomes sizing.