# Reducing Bet Placement Latency for a Sportsbook

> TechSupport Team supplied dedicated L2/L3 engineers for OddsForge’s Java and Kafka sportsbook. In a four-week pilot, the client’s Datadog monitoring recorded p95 bet-placement latency falling from 840 ms to 244 ms in the measured tournament window.

- Publisher: TechSupport Team
- Canonical page: https://techsupport-team.com/case-studies/sportsbook-bet-placement-latency/
- Updated: 2026-11-16

Client: OddsForge · iGaming & Betting · Maltese online sportsbook

## Client and engagement facts

- **Client:** OddsForge
- **Industry:** iGaming & Betting
- **Business context:** Maltese online sportsbook
- **Delivery model:** L2/L3 staff augmentation
- **Application:** Bet-placement API with Kafka pipeline and settlement workers
- **Supplied geography:** Maltese

## Application stack and tools

### Application stack

- Java/Spring
- Kafka
- Redis
- PostgreSQL
- EKS

### Tools used

- Datadog
- PagerDuty
- Grafana
- Jira

Related services: [Hire support engineers](https://techsupport-team.com/hire/support-engineers/)

## Why Support Was Needed

During tournaments, bet requests competed with settlement workers for the same Kafka partitions. The sportsbook used staff augmentation to keep incident command and production approval with its own team.

## Technical Work

- Reproduced the peak slowdown in staging with a 20-minute load test.
- Identified partition skew, long poll intervals and garbage collection pauses.
- Rebalanced partitions and added a partition-level Kafka lag alert in Grafana.

## Representative Incident

During a quarter-final, one Kafka partition drew 38% of traffic after a consumer restarted with stale metadata. The on-call engineer paused the consumer and rejoined the group. Flow recovered in 11 minutes, and the team added a check 30 minutes before kickoff.

## Results and how they were measured

Table: Reported results and their measurement scope

| Metric | Baseline | Result | Units | Measurement window | Source and scope |
| --- | --- | --- | --- | --- | --- |
| p95 bet-placement latency | 840 ms | 244 ms | ms | 4-week pilot | Client Datadog |
| Peak-week incidents | 14 | 2 |  | 4 weeks | Client PagerDuty; Scope: The four-week pilot excluded two vendor-feed incidents from the platform incident count. |
| Kafka lag alerts | 31/wk | 4/wk | /wk | 4 weeks | Client Grafana |

## Evidence and Limitations

The four-week pilot excluded two vendor-feed incidents from the platform incident count. Tournament volumes varied. The peak-week comparison uses the quarter-final window.

## Onboarding and Change Approval

Onboarding took seven business days, including the client interview. The engineers received least-privilege read access, and OddsForge approved production changes through Jira.

## Scope a Similar Engagement

Describe the API, the peak window and the incident class your lead wants a dedicated engineer to own. [Request a support plan](https://techsupport-team.com/contact/) or explore [dedicated support engineers](https://techsupport-team.com/hire/support-engineers/).

## Related pages

- [Hire support engineers](https://techsupport-team.com/hire/support-engineers/): The related service for this engagement.

## Contact

Request engineers from TechSupport Team or ask a question: https://techsupport-team.com/contact/
