Game Performance Index Overview

• Game Performance Index (GPI) is a composite metric that quantifies gaming performance by integrating visual-motor tracking and mobile device evaluation.
• It aggregates raw data such as bits per second from visual changes and normalized sub-indices like FPS, battery, and responsiveness into a unified score.
• GPI informs benchmarks for esports talent and device suitability, enabling targeted optimizations and comparative analyses.

The Game Performance Index (GPI) is a quantitative metric designed to aggregate multifaceted aspects of performance in gaming contexts—from cognitive information processing under visually complex conditions to mobile device suitability for high-fidelity interactive entertainment. While independently originated in cognitive benchmarking (Bátfai et al., 2018) and mobile device evaluation (Dar et al., 2019), the GPI label in each context denotes a rigorously defined, operationalized score summarizing user-centered or user-experience relevant game performance.

1. Cognitive Benchmarking: GPI in the BrainB Series

In Batfai et al.'s BrainB Series 6 benchmark (Bátfai et al., 2018), the Game Performance Index quantifies a participant’s capacity for visual-motor tracking under dynamically titrated visual complexity. Complexity, $C$, is operationalized as the change rate in the display region surrounding a moving “character” (target box): specifically, the number of RGB-changed bits, $\Delta B$, between consecutive 100 ms frames, normalized to bits per second: $C(t) = \Delta B/\Delta t$. Instantaneous $C$ values are computed via a loop that samples every 100 ms.

Participants direct a mouse pointer to follow the target. The squared Euclidean distance between pointer and target center is checked each cycle. If a subject remains off-target ($\textrm{dist} > 121$ pixels) for over 1.2 s (12 cycles), a “loss” event is recorded; visual complexity is decreased (fewer distractors, slower moving objects). Conversely, “found” events (on-target for 1.2 s after being lost) trigger a complexity increase. The lists of complexity values at loss (“found$\rightarrow$lost”) and recovery (“lost$\rightarrow$found”) events are maintained in vectors, which, by design, alternate in time.

At the conclusion of a 10-minute session, two means are computed: $$m_1 = \textrm{mean}(\textrm{lost}\rightarrow\textrm{found})$$, $$m_2 = \textrm{mean}(\textrm{found}\rightarrow\textrm{lost})$$. The raw GPI is the average,

$\mathrm{GPI} = \frac{1}{2}(m_1 + m_2),$

then scaled from bits per second to Kilobytes via division by $\Delta B$0: $\Delta B$1

A higher GPI indicates a subject can sustain fluid tracking at higher ambient visual complexity. Empirical results indicate that esports athletes in the DEAC-Hackers group (mean GPI = 3.71 KB) underperformed in visual complexity tolerance compared to undergraduate programmers (mean GPI = 4.95 KB). The internal validity is supported by within-subject logging of means and variances for both “lost” and “found” event complexity, and the consistent observation that mean complexity at loss events exceeds that at recovery, i.e., $\Delta B$2.

2. Mobile Device Assessment: GPI as a Composite User Experience Measure

In mobile device contexts, particularly as implemented at Samsung Electronics (Dar et al., 2019), the Game Performance Index is constructed as a composite, weighted sum of normalized sub-indices reflecting both perceptual and system performance dimensions during game play. The conceptual framework consists of a three-stage mapping pipeline:

1. Metric $\Delta B$3 Sub-Index: Raw device metrics (e.g., average FPS, battery drain rate, surface temperature) are mapped to sub-index scores ($\Delta B$4) via empirically or expert-defined normalization curves.
2. Sub-Index $\Delta B$5 Main-Index: Perceptual categories (visual smoothness, graphics quality, battery, temperature, swiftness, responsiveness) aggregate sub-index scores using weights $\Delta B$6, yielding categorical main-index values $\Delta B$7.
3. Main-Index $\Delta B$8 GPI: The GPI itself is aggregated as:

$\Delta B$9

$C(t) = \Delta B/\Delta t$0

with $C(t) = \Delta B/\Delta t$1 denoting the j-th raw metric in category $C(t) = \Delta B/\Delta t$2, $C(t) = \Delta B/\Delta t$3 the normalization, $C(t) = \Delta B/\Delta t$4 and $C(t) = \Delta B/\Delta t$5 intra/inter-category weights.

Distinct weighting sets, $C(t) = \Delta B/\Delta t$6, are defined for competitive (emphasizing frame rate and latency) versus casual (emphasizing battery and thermal comfort) gamer profiles. For example, $C(t) = \Delta B/\Delta t$7, $C(t) = \Delta B/\Delta t$8 for competitive, and $C(t) = \Delta B/\Delta t$9, $C$0 for casual. The specific weights in Samsung’s production systems are proprietary.

3. Performance Metrics and Experimental Outcomes

In both benchmarking paradigms, the choice of input metrics and event triggers is central:

• BrainB Series 6: Pixel-level complexity (bits per second), determined directly by visual change rate in the target’s immediate display region, modulated by user’s ability to keep the pointer on target (Bátfai et al., 2018).
• Samsung GPI: Six main categories—visual smoothness (average and 1%-percentile FPS), graphical quality (texture/effects settings), battery ($C$1 drain/hour), temperature (peak, throttling onset), swiftness (application and scene load times), responsiveness (touch/input latency in ms)—collectively drive the final index (Dar et al., 2019).

In the Samsung implementation, device testing (nine 2018–2019 flagship models, anonymized A–I) involved controlled environment sessions with system telemetry and high-speed input latency instrumentation. Empirical GPI values ranged from 75–86 (Competitive) and 70–86 (Casual) across devices, with within-device variation $C$21.5 points and cross-device differences $C$32 points significant at $C$4. Top-tier performance for “competitive” use was defined as GPI $C$5, while mid-range or older hardware corresponded to GPI $C$6 75.

4. Interpretation, Validity, and Use Cases

Within cognitive benchmarking, the GPI is framed as a putative measure of the visual-motor system’s information processing “speed” or “capacity,” analogous in spirit to classical vigilance tasks. The operationalization via bits per second (scaled to kilobytes) is intended as an objective, repeatable marker of dynamic attention resilience in distractor-rich environments. A higher GPI corresponds to greater ability to tolerate—and recover tracking within—more visually complex conditions. The authors propose potential use in esports talent identification, complementing conventional reaction time and memory batteries (Bátfai et al., 2018).

For device assessment, the GPI provides both a consumer-facing “one-number shorthand” for relative ranking and a developer guideline for setting minimum device support levels or automatic graphics presets. For instance, competitive GPI $C$7 is recommended for fast-twitch apps, while casual game development targets devices with casual GPI $C$8. Category-level sub-scores can inform fine-grained optimizations.

5. Methodological Limitations and Open Questions

The cognitive GPI (BrainB) shows platform sensitivity; results may not generalize outside tightly controlled software/hardware/mouse settings. The system’s lack of timestamped lose/found event logging in Series 6 precludes analysis of reaction times under Hick’s law paradigms, though this is proposed for Series 7. Generalizability beyond single-institution samples and evidence of construct validity against established neuropsychological batteries remain unresolved (Bátfai et al., 2018).

The mobile device GPI’s weights are driven by expert choice and may not reflect user-perceived priorities outside the tested demographic. Current implementations lack network performance metrics (lag, packet loss), potentially limiting applicability to multiplayer or cloud gaming contexts. Only three flagship titles were profiled, leaving generalization to lighter games or VR/AR unassessed. Extensions could include machine-learned normalization from user surveys, inclusion of network QoS, and adaptation to new form factors such as foldables (Dar et al., 2019).

6. Comparative Table of GPI Implementations

Context	Input Metrics / Categories	GPI Calculation
BrainB Series 6 (Bátfai et al., 2018)	Bit-rate (changed pixels/sec); “lose” and “find” events	$C$9 [KB]
Samsung Mobile (Dar et al., 2019)	FPS, graphics, battery, temp, swiftness, responsiveness	$\textrm{dist} > 121$0; $\textrm{dist} > 121$1 [0–100 scale]

The table demonstrates the divergence in domain and aggregation methodology despite the shared “GPI” nomenclature. The BrainB GPI yields a subject’s information processing capacity in kilobytes, while the Samsung GPI produces a normalized, weighted sum for comparative device rating.

7. Prospective Applications and Future Development

Proposed future directions for the cognitive GPI include integration into standardized psychological testing regimes and inclusion of timestamped event data for reaction-time analysis. Validation across diverse cohorts and with established cognitive assessment tools is recognized as essential.

For device-oriented GPI, the roadmap favors broadening the metric to incorporate network conditions, leveraging user-driven mapping calibration, and cross-platform standardization for industry-wide uptake. Expanding to VR/AR environments and new device modalities is also highlighted.

Both instantiations of GPI exemplify the trend toward composite, user-experience-centric metrics in both the cognitive neurosciences and device engineering, balancing empirical measurement with practical interpretability for broad academic and industry audiences.