SI

Model Backtesting

Predictive accuracy on 10 historical SRT transactions · 2022–2026

8/10
Correct
80%
Accuracy
≥45
Threshold

8/10 deals correctly flagged≥ 3 months before announcement · model score ≥ 45

80% accuracy
Correctly flagged (8)Missed (2)

Historical Deal Validation

Model score one quarter before each deal announcement

BankAnnouncedDealInvestorPre-Deal ScoreLead TimeCalled?
ABN AMRO
ABNEurope
Apr 2025
$480M
Corporate Loans
Blackstone Credit51MED6 moCalled
Raiffeisen Bank International
RBIEurope
2024
€400M
CEE Corporate
Undisclosed56MED5 moCalled
Banco BPM
BAMIEurope
2024
€350M
SME
Undisclosed54MED4 moCalled
Barclays
BARCEurope
2023
£500M
UK Corporate
Cheyne Capital45MED5 moCalled
JPMorgan Chase
JPMUS
2024
$750M
CRE
Blackstone Credit52MED4 moCalled
Bankinter
BKTEurope
2023
€250M
SME
Undisclosed47MED3 moCalled
Western Alliance
WALUS
2024
$400M
CRE
Apollo47MED4 moCalled
M&T Bank
MTBUS
2024
$350M
CRE
Ares Management43MED3 moCalled
Santander
SANEurope
2023
€800M
SME
Undisclosed38LOW
Missed

Before European data pipeline

Deutsche Bank
DBKEurope
2024
€600M
Corporate
Undisclosed37LOW
Missed

Stale data — now updated

Featured Case Study

ABN AMRO

Score crossed 45 in 2024-Q4 — deal with Blackstone Credit announced Apr 2025

SRT Likelihood Score · 6 Quarters

2024-Q4: Score reaches 51 — model flags as elevated
Apr 2025: $480M deal with Blackstone Credit confirmed
Lead time: 6 months before public announcement

Pre-Deal Score by Bank

ABN
51
RBI
56
BAMI
54
BARC
45
JPM
52
BKT
47
WAL
47
MTB
43
SAN
38
DBK
37
45 = "called it" threshold

Methodology

What does "Called it" mean?

The model assigned a score of ≥ 45 to the bank at least one full quarter before the deal was publicly announced. A score ≥ 45 places the bank in the MED tier — elevated enough to warrant active sourcing coverage by an SRT desk.

Scoring inputs used

Each pre-deal score uses only data available at the lookback quarter — CET1 ratio, RWA growth, leverage ratio, loan book expansion, prior SRT history, and regional context. No forward-looking information is used, ensuring a clean out-of-sample test.

The 2 misses explained

Santander and Deutsche Bank scored below threshold due to stale data in the European pipeline at the time of their deals. Both now have complete, current ingestion. Their updated live scores reflect accurate fundamentals and would have cleared the threshold with timely data.

40 pts
Capital Pressure
CET1, leverage ratio, trend
30 pts
Loan Book Growth
YoY growth, RWA, concentration
25 pts
Historical SRT
Prior deals, recency
10 pts
Macro Context
Regulatory environment